Add files using upload-large-folder tool

adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-1000/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-2000/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-2000/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-2000/latest

@@ -0,0 +1 @@
+global_step2000

checkpoint-2000/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-2000/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-2000/trainer_state.json

@@ -0,0 +1,1457 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.697072549851506,
+  "eval_steps": 600,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00848536274925753,
+      "grad_norm": 0.4898678891363344,
+      "learning_rate": 8.488964346349746e-07,
+      "loss": 1.8056,
+      "step": 10
+    },
+    {
+      "epoch": 0.01697072549851506,
+      "grad_norm": 0.3537473179717183,
+      "learning_rate": 1.6977928692699491e-06,
+      "loss": 1.7621,
+      "step": 20
+    },
+    {
+      "epoch": 0.025456088247772592,
+      "grad_norm": 0.28215953004159977,
+      "learning_rate": 2.546689303904924e-06,
+      "loss": 1.7571,
+      "step": 30
+    },
+    {
+      "epoch": 0.03394145099703012,
+      "grad_norm": 0.27446565146764923,
+      "learning_rate": 3.3955857385398982e-06,
+      "loss": 1.7136,
+      "step": 40
+    },
+    {
+      "epoch": 0.04242681374628765,
+      "grad_norm": 0.17051549768176558,
+      "learning_rate": 4.244482173174873e-06,
+      "loss": 1.6767,
+      "step": 50
+    },
+    {
+      "epoch": 0.050912176495545185,
+      "grad_norm": 0.17763882467320422,
+      "learning_rate": 5.093378607809848e-06,
+      "loss": 1.6371,
+      "step": 60
+    },
+    {
+      "epoch": 0.05939753924480271,
+      "grad_norm": 0.14311462596290048,
+      "learning_rate": 5.942275042444822e-06,
+      "loss": 1.6324,
+      "step": 70
+    },
+    {
+      "epoch": 0.06788290199406025,
+      "grad_norm": 0.1659540846071645,
+      "learning_rate": 6.7911714770797965e-06,
+      "loss": 1.6062,
+      "step": 80
+    },
+    {
+      "epoch": 0.07636826474331777,
+      "grad_norm": 0.20064072815620043,
+      "learning_rate": 7.640067911714771e-06,
+      "loss": 1.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0848536274925753,
+      "grad_norm": 0.2179045681711979,
+      "learning_rate": 8.488964346349745e-06,
+      "loss": 1.5898,
+      "step": 100
+    },
+    {
+      "epoch": 0.09333899024183284,
+      "grad_norm": 0.23866012053128668,
+      "learning_rate": 9.337860780984721e-06,
+      "loss": 1.5924,
+      "step": 110
+    },
+    {
+      "epoch": 0.10182435299109037,
+      "grad_norm": 0.18578051776430282,
+      "learning_rate": 1.0186757215619695e-05,
+      "loss": 1.5877,
+      "step": 120
+    },
+    {
+      "epoch": 0.1103097157403479,
+      "grad_norm": 0.2216509707409362,
+      "learning_rate": 1.103565365025467e-05,
+      "loss": 1.5947,
+      "step": 130
+    },
+    {
+      "epoch": 0.11879507848960542,
+      "grad_norm": 0.20427142255694086,
+      "learning_rate": 1.1884550084889643e-05,
+      "loss": 1.5841,
+      "step": 140
+    },
+    {
+      "epoch": 0.12728044123886295,
+      "grad_norm": 0.1765851415675038,
+      "learning_rate": 1.2733446519524619e-05,
+      "loss": 1.5878,
+      "step": 150
+    },
+    {
+      "epoch": 0.1357658039881205,
+      "grad_norm": 0.1769355117060811,
+      "learning_rate": 1.3582342954159593e-05,
+      "loss": 1.5795,
+      "step": 160
+    },
+    {
+      "epoch": 0.14425116673737803,
+      "grad_norm": 0.1617675663096666,
+      "learning_rate": 1.4431239388794569e-05,
+      "loss": 1.5549,
+      "step": 170
+    },
+    {
+      "epoch": 0.15273652948663555,
+      "grad_norm": 0.17302259072151574,
+      "learning_rate": 1.5280135823429543e-05,
+      "loss": 1.5808,
+      "step": 180
+    },
+    {
+      "epoch": 0.1612218922358931,
+      "grad_norm": 0.16876039012432806,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 1.5676,
+      "step": 190
+    },
+    {
+      "epoch": 0.1697072549851506,
+      "grad_norm": 0.19627360154037596,
+      "learning_rate": 1.697792869269949e-05,
+      "loss": 1.5598,
+      "step": 200
+    },
+    {
+      "epoch": 0.17819261773440814,
+      "grad_norm": 0.16078510362361015,
+      "learning_rate": 1.7826825127334465e-05,
+      "loss": 1.5667,
+      "step": 210
+    },
+    {
+      "epoch": 0.18667798048366568,
+      "grad_norm": 0.16044786518959703,
+      "learning_rate": 1.8675721561969442e-05,
+      "loss": 1.5815,
+      "step": 220
+    },
+    {
+      "epoch": 0.1951633432329232,
+      "grad_norm": 0.15656958873834717,
+      "learning_rate": 1.9524617996604416e-05,
+      "loss": 1.5576,
+      "step": 230
+    },
+    {
+      "epoch": 0.20364870598218074,
+      "grad_norm": 0.1687290471357602,
+      "learning_rate": 2.037351443123939e-05,
+      "loss": 1.5453,
+      "step": 240
+    },
+    {
+      "epoch": 0.21213406873143828,
+      "grad_norm": 0.1519017348276184,
+      "learning_rate": 2.1222410865874364e-05,
+      "loss": 1.5554,
+      "step": 250
+    },
+    {
+      "epoch": 0.2206194314806958,
+      "grad_norm": 0.15761892005160086,
+      "learning_rate": 2.207130730050934e-05,
+      "loss": 1.5494,
+      "step": 260
+    },
+    {
+      "epoch": 0.22910479422995333,
+      "grad_norm": 0.16857088482977495,
+      "learning_rate": 2.2920203735144312e-05,
+      "loss": 1.5794,
+      "step": 270
+    },
+    {
+      "epoch": 0.23759015697921085,
+      "grad_norm": 0.1678705209913503,
+      "learning_rate": 2.3769100169779286e-05,
+      "loss": 1.5373,
+      "step": 280
+    },
+    {
+      "epoch": 0.2460755197284684,
+      "grad_norm": 0.14812649566587394,
+      "learning_rate": 2.461799660441426e-05,
+      "loss": 1.5504,
+      "step": 290
+    },
+    {
+      "epoch": 0.2545608824777259,
+      "grad_norm": 0.17651916734325857,
+      "learning_rate": 2.5466893039049238e-05,
+      "loss": 1.5607,
+      "step": 300
+    },
+    {
+      "epoch": 0.26304624522698344,
+      "grad_norm": 0.14883055338507856,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 1.5311,
+      "step": 310
+    },
+    {
+      "epoch": 0.271531607976241,
+      "grad_norm": 0.15787522753231265,
+      "learning_rate": 2.7164685908319186e-05,
+      "loss": 1.5656,
+      "step": 320
+    },
+    {
+      "epoch": 0.2800169707254985,
+      "grad_norm": 0.1625232940237689,
+      "learning_rate": 2.801358234295416e-05,
+      "loss": 1.5686,
+      "step": 330
+    },
+    {
+      "epoch": 0.28850233347475607,
+      "grad_norm": 0.18505951289343867,
+      "learning_rate": 2.8862478777589137e-05,
+      "loss": 1.5474,
+      "step": 340
+    },
+    {
+      "epoch": 0.29698769622401355,
+      "grad_norm": 0.13785772316349984,
+      "learning_rate": 2.9711375212224108e-05,
+      "loss": 1.5696,
+      "step": 350
+    },
+    {
+      "epoch": 0.3054730589732711,
+      "grad_norm": 0.13531274658248552,
+      "learning_rate": 3.0560271646859086e-05,
+      "loss": 1.5551,
+      "step": 360
+    },
+    {
+      "epoch": 0.31395842172252864,
+      "grad_norm": 0.1366381415368909,
+      "learning_rate": 3.140916808149406e-05,
+      "loss": 1.524,
+      "step": 370
+    },
+    {
+      "epoch": 0.3224437844717862,
+      "grad_norm": 0.14587220569353926,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 1.5515,
+      "step": 380
+    },
+    {
+      "epoch": 0.3309291472210437,
+      "grad_norm": 0.13336349383744864,
+      "learning_rate": 3.310696095076401e-05,
+      "loss": 1.5457,
+      "step": 390
+    },
+    {
+      "epoch": 0.3394145099703012,
+      "grad_norm": 0.1772016947970983,
+      "learning_rate": 3.395585738539898e-05,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 0.34789987271955874,
+      "grad_norm": 0.13819420575084573,
+      "learning_rate": 3.4804753820033956e-05,
+      "loss": 1.5326,
+      "step": 410
+    },
+    {
+      "epoch": 0.3563852354688163,
+      "grad_norm": 0.12729862167862188,
+      "learning_rate": 3.565365025466893e-05,
+      "loss": 1.5387,
+      "step": 420
+    },
+    {
+      "epoch": 0.3648705982180738,
+      "grad_norm": 0.11777082851399363,
+      "learning_rate": 3.6502546689303904e-05,
+      "loss": 1.5587,
+      "step": 430
+    },
+    {
+      "epoch": 0.37335596096733137,
+      "grad_norm": 0.15372268131323022,
+      "learning_rate": 3.7351443123938885e-05,
+      "loss": 1.5362,
+      "step": 440
+    },
+    {
+      "epoch": 0.3818413237165889,
+      "grad_norm": 0.12616185572252248,
+      "learning_rate": 3.820033955857386e-05,
+      "loss": 1.5548,
+      "step": 450
+    },
+    {
+      "epoch": 0.3903266864658464,
+      "grad_norm": 0.1311200786303391,
+      "learning_rate": 3.904923599320883e-05,
+      "loss": 1.5409,
+      "step": 460
+    },
+    {
+      "epoch": 0.39881204921510394,
+      "grad_norm": 0.1707919112561785,
+      "learning_rate": 3.989813242784381e-05,
+      "loss": 1.5509,
+      "step": 470
+    },
+    {
+      "epoch": 0.4072974119643615,
+      "grad_norm": 0.14660149264284913,
+      "learning_rate": 4.074702886247878e-05,
+      "loss": 1.5433,
+      "step": 480
+    },
+    {
+      "epoch": 0.415782774713619,
+      "grad_norm": 0.12478895483834351,
+      "learning_rate": 4.1595925297113755e-05,
+      "loss": 1.5382,
+      "step": 490
+    },
+    {
+      "epoch": 0.42426813746287656,
+      "grad_norm": 0.12327957445795817,
+      "learning_rate": 4.244482173174873e-05,
+      "loss": 1.5515,
+      "step": 500
+    },
+    {
+      "epoch": 0.43275350021213405,
+      "grad_norm": 0.12922777738650987,
+      "learning_rate": 4.32937181663837e-05,
+      "loss": 1.5688,
+      "step": 510
+    },
+    {
+      "epoch": 0.4412388629613916,
+      "grad_norm": 0.12486802189783415,
+      "learning_rate": 4.414261460101868e-05,
+      "loss": 1.5452,
+      "step": 520
+    },
+    {
+      "epoch": 0.44972422571064913,
+      "grad_norm": 0.1360610874577123,
+      "learning_rate": 4.499151103565366e-05,
+      "loss": 1.5493,
+      "step": 530
+    },
+    {
+      "epoch": 0.45820958845990667,
+      "grad_norm": 0.1884897685356775,
+      "learning_rate": 4.5840407470288625e-05,
+      "loss": 1.5511,
+      "step": 540
+    },
+    {
+      "epoch": 0.4666949512091642,
+      "grad_norm": 0.12446302384809525,
+      "learning_rate": 4.6689303904923606e-05,
+      "loss": 1.5458,
+      "step": 550
+    },
+    {
+      "epoch": 0.4751803139584217,
+      "grad_norm": 0.13169591804768588,
+      "learning_rate": 4.753820033955857e-05,
+      "loss": 1.5569,
+      "step": 560
+    },
+    {
+      "epoch": 0.48366567670767924,
+      "grad_norm": 0.1343809247449631,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 1.5408,
+      "step": 570
+    },
+    {
+      "epoch": 0.4921510394569368,
+      "grad_norm": 0.14024589853602,
+      "learning_rate": 4.923599320882852e-05,
+      "loss": 1.5487,
+      "step": 580
+    },
+    {
+      "epoch": 0.5006364022061943,
+      "grad_norm": 0.16240429253875313,
+      "learning_rate": 4.999999560970061e-05,
+      "loss": 1.5488,
+      "step": 590
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "grad_norm": 0.12575424857894482,
+      "learning_rate": 4.999946877563971e-05,
+      "loss": 1.532,
+      "step": 600
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "eval_loss": 1.519254446029663,
+      "eval_runtime": 53.3242,
+      "eval_samples_per_second": 7.145,
+      "eval_steps_per_second": 0.9,
+      "step": 600
+    },
+    {
+      "epoch": 0.5176071277047094,
+      "grad_norm": 0.18688482756329736,
+      "learning_rate": 4.999806390290309e-05,
+      "loss": 1.5544,
+      "step": 610
+    },
+    {
+      "epoch": 0.5260924904539669,
+      "grad_norm": 0.12425469431830571,
+      "learning_rate": 4.999578104083307e-05,
+      "loss": 1.5443,
+      "step": 620
+    },
+    {
+      "epoch": 0.5345778532032245,
+      "grad_norm": 0.1299027485420099,
+      "learning_rate": 4.999262026960902e-05,
+      "loss": 1.5569,
+      "step": 630
+    },
+    {
+      "epoch": 0.543063215952482,
+      "grad_norm": 0.11441754852508934,
+      "learning_rate": 4.998858170024449e-05,
+      "loss": 1.5316,
+      "step": 640
+    },
+    {
+      "epoch": 0.5515485787017395,
+      "grad_norm": 0.14888547248976478,
+      "learning_rate": 4.998366547458326e-05,
+      "loss": 1.5177,
+      "step": 650
+    },
+    {
+      "epoch": 0.560033941450997,
+      "grad_norm": 0.14859292774768867,
+      "learning_rate": 4.997787176529449e-05,
+      "loss": 1.5394,
+      "step": 660
+    },
+    {
+      "epoch": 0.5685193042002545,
+      "grad_norm": 0.12499154376539734,
+      "learning_rate": 4.997120077586651e-05,
+      "loss": 1.5554,
+      "step": 670
+    },
+    {
+      "epoch": 0.5770046669495121,
+      "grad_norm": 0.1218974898058821,
+      "learning_rate": 4.9963652740599774e-05,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.5854900296987696,
+      "grad_norm": 0.1273110498715124,
+      "learning_rate": 4.995522792459859e-05,
+      "loss": 1.5349,
+      "step": 690
+    },
+    {
+      "epoch": 0.5939753924480271,
+      "grad_norm": 0.12115412881719101,
+      "learning_rate": 4.994592662376183e-05,
+      "loss": 1.5419,
+      "step": 700
+    },
+    {
+      "epoch": 0.6024607551972847,
+      "grad_norm": 0.14855096330233286,
+      "learning_rate": 4.99357491647725e-05,
+      "loss": 1.513,
+      "step": 710
+    },
+    {
+      "epoch": 0.6109461179465422,
+      "grad_norm": 0.11407988659327956,
+      "learning_rate": 4.992469590508628e-05,
+      "loss": 1.5243,
+      "step": 720
+    },
+    {
+      "epoch": 0.6194314806957998,
+      "grad_norm": 0.1197712643781127,
+      "learning_rate": 4.9912767232919035e-05,
+      "loss": 1.5177,
+      "step": 730
+    },
+    {
+      "epoch": 0.6279168434450573,
+      "grad_norm": 0.12400515877262065,
+      "learning_rate": 4.9899963567233074e-05,
+      "loss": 1.5619,
+      "step": 740
+    },
+    {
+      "epoch": 0.6364022061943148,
+      "grad_norm": 0.12250385257708406,
+      "learning_rate": 4.988628535772249e-05,
+      "loss": 1.539,
+      "step": 750
+    },
+    {
+      "epoch": 0.6448875689435724,
+      "grad_norm": 0.1262441090496857,
+      "learning_rate": 4.987173308479738e-05,
+      "loss": 1.5195,
+      "step": 760
+    },
+    {
+      "epoch": 0.6533729316928298,
+      "grad_norm": 0.12459694416473029,
+      "learning_rate": 4.985630725956694e-05,
+      "loss": 1.5462,
+      "step": 770
+    },
+    {
+      "epoch": 0.6618582944420874,
+      "grad_norm": 0.12985189006106762,
+      "learning_rate": 4.9840008423821527e-05,
+      "loss": 1.5113,
+      "step": 780
+    },
+    {
+      "epoch": 0.6703436571913449,
+      "grad_norm": 0.12689306141471304,
+      "learning_rate": 4.9822837150013636e-05,
+      "loss": 1.5201,
+      "step": 790
+    },
+    {
+      "epoch": 0.6788290199406024,
+      "grad_norm": 0.15393156370587963,
+      "learning_rate": 4.980479404123778e-05,
+      "loss": 1.5121,
+      "step": 800
+    },
+    {
+      "epoch": 0.68731438268986,
+      "grad_norm": 0.13213701895207608,
+      "learning_rate": 4.978587973120931e-05,
+      "loss": 1.5307,
+      "step": 810
+    },
+    {
+      "epoch": 0.6957997454391175,
+      "grad_norm": 0.11561354931316294,
+      "learning_rate": 4.9766094884242184e-05,
+      "loss": 1.5316,
+      "step": 820
+    },
+    {
+      "epoch": 0.7042851081883751,
+      "grad_norm": 0.12414772399330044,
+      "learning_rate": 4.974544019522559e-05,
+      "loss": 1.5148,
+      "step": 830
+    },
+    {
+      "epoch": 0.7127704709376326,
+      "grad_norm": 0.1171652849153521,
+      "learning_rate": 4.972391638959959e-05,
+      "loss": 1.5096,
+      "step": 840
+    },
+    {
+      "epoch": 0.7212558336868902,
+      "grad_norm": 0.12868937349582316,
+      "learning_rate": 4.9701524223329585e-05,
+      "loss": 1.5282,
+      "step": 850
+    },
+    {
+      "epoch": 0.7297411964361477,
+      "grad_norm": 0.1200015077117309,
+      "learning_rate": 4.967826448287981e-05,
+      "loss": 1.5512,
+      "step": 860
+    },
+    {
+      "epoch": 0.7382265591854051,
+      "grad_norm": 0.12340885660045105,
+      "learning_rate": 4.96541379851857e-05,
+      "loss": 1.5394,
+      "step": 870
+    },
+    {
+      "epoch": 0.7467119219346627,
+      "grad_norm": 0.12976937691467555,
+      "learning_rate": 4.962914557762517e-05,
+      "loss": 1.51,
+      "step": 880
+    },
+    {
+      "epoch": 0.7551972846839202,
+      "grad_norm": 0.11912878476038466,
+      "learning_rate": 4.9603288137988905e-05,
+      "loss": 1.5294,
+      "step": 890
+    },
+    {
+      "epoch": 0.7636826474331778,
+      "grad_norm": 0.1299625480337927,
+      "learning_rate": 4.957656657444947e-05,
+      "loss": 1.507,
+      "step": 900
+    },
+    {
+      "epoch": 0.7721680101824353,
+      "grad_norm": 0.12380144459698468,
+      "learning_rate": 4.954898182552946e-05,
+      "loss": 1.5376,
+      "step": 910
+    },
+    {
+      "epoch": 0.7806533729316928,
+      "grad_norm": 0.13139339643682763,
+      "learning_rate": 4.9520534860068535e-05,
+      "loss": 1.5291,
+      "step": 920
+    },
+    {
+      "epoch": 0.7891387356809504,
+      "grad_norm": 0.13088956203983898,
+      "learning_rate": 4.949122667718935e-05,
+      "loss": 1.5239,
+      "step": 930
+    },
+    {
+      "epoch": 0.7976240984302079,
+      "grad_norm": 0.12586052988453703,
+      "learning_rate": 4.94610583062625e-05,
+      "loss": 1.5525,
+      "step": 940
+    },
+    {
+      "epoch": 0.8061094611794655,
+      "grad_norm": 0.12020996031652877,
+      "learning_rate": 4.943003080687035e-05,
+      "loss": 1.5525,
+      "step": 950
+    },
+    {
+      "epoch": 0.814594823928723,
+      "grad_norm": 0.12866375954060869,
+      "learning_rate": 4.9398145268769856e-05,
+      "loss": 1.5266,
+      "step": 960
+    },
+    {
+      "epoch": 0.8230801866779804,
+      "grad_norm": 0.13166136756817035,
+      "learning_rate": 4.936540281185423e-05,
+      "loss": 1.5041,
+      "step": 970
+    },
+    {
+      "epoch": 0.831565549427238,
+      "grad_norm": 0.12481946698483787,
+      "learning_rate": 4.933180458611364e-05,
+      "loss": 1.5271,
+      "step": 980
+    },
+    {
+      "epoch": 0.8400509121764955,
+      "grad_norm": 0.12264463761209114,
+      "learning_rate": 4.9297351771594844e-05,
+      "loss": 1.5354,
+      "step": 990
+    },
+    {
+      "epoch": 0.8485362749257531,
+      "grad_norm": 0.11985452856537594,
+      "learning_rate": 4.926204557835968e-05,
+      "loss": 1.5167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8570216376750106,
+      "grad_norm": 0.13125396521190327,
+      "learning_rate": 4.9225887246442634e-05,
+      "loss": 1.5282,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8655070004242681,
+      "grad_norm": 0.12730192328072554,
+      "learning_rate": 4.918887804580725e-05,
+      "loss": 1.5089,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8739923631735257,
+      "grad_norm": 0.12724644219344786,
+      "learning_rate": 4.915101927630153e-05,
+      "loss": 1.4964,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8824777259227832,
+      "grad_norm": 0.13578611501833232,
+      "learning_rate": 4.911231226761227e-05,
+      "loss": 1.5189,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8909630886720408,
+      "grad_norm": 0.13577513964986457,
+      "learning_rate": 4.90727583792184e-05,
+      "loss": 1.5149,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8994484514212983,
+      "grad_norm": 0.1269735011676505,
+      "learning_rate": 4.903235900034317e-05,
+      "loss": 1.5066,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9079338141705557,
+      "grad_norm": 0.13250058214235566,
+      "learning_rate": 4.899111554990543e-05,
+      "loss": 1.5129,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9164191769198133,
+      "grad_norm": 0.13130735246433495,
+      "learning_rate": 4.894902947646975e-05,
+      "loss": 1.5156,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9249045396690708,
+      "grad_norm": 0.1273580180253049,
+      "learning_rate": 4.890610225819553e-05,
+      "loss": 1.5324,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9333899024183284,
+      "grad_norm": 0.13155314243939242,
+      "learning_rate": 4.8862335402785136e-05,
+      "loss": 1.5106,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9418752651675859,
+      "grad_norm": 0.13564895211984299,
+      "learning_rate": 4.88177304474309e-05,
+      "loss": 1.5067,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9503606279168434,
+      "grad_norm": 0.12774735587114736,
+      "learning_rate": 4.877228895876115e-05,
+      "loss": 1.5182,
+      "step": 1120
+    },
+    {
+      "epoch": 0.958845990666101,
+      "grad_norm": 0.1307997709537685,
+      "learning_rate": 4.872601253278517e-05,
+      "loss": 1.4969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9673313534153585,
+      "grad_norm": 0.1304794845040634,
+      "learning_rate": 4.867890279483717e-05,
+      "loss": 1.5264,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9758167161646161,
+      "grad_norm": 0.13666141796489684,
+      "learning_rate": 4.8630961399519206e-05,
+      "loss": 1.5467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9843020789138736,
+      "grad_norm": 0.1370278303190263,
+      "learning_rate": 4.8582190030643e-05,
+      "loss": 1.5127,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9927874416631312,
+      "grad_norm": 0.1390936629299565,
+      "learning_rate": 4.8532590401170894e-05,
+      "loss": 1.5058,
+      "step": 1170
+    },
+    {
+      "epoch": 1.0012728044123886,
+      "grad_norm": 0.12934475548108287,
+      "learning_rate": 4.848216425315561e-05,
+      "loss": 1.5202,
+      "step": 1180
+    },
+    {
+      "epoch": 1.0097581671616462,
+      "grad_norm": 0.13898591683370803,
+      "learning_rate": 4.843091335767913e-05,
+      "loss": 1.4563,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "grad_norm": 0.17488231535826249,
+      "learning_rate": 4.837883951479043e-05,
+      "loss": 1.4402,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "eval_loss": 1.4955657720565796,
+      "eval_runtime": 52.424,
+      "eval_samples_per_second": 7.268,
+      "eval_steps_per_second": 0.916,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0267288926601612,
+      "grad_norm": 0.1536036344095855,
+      "learning_rate": 4.832594455344229e-05,
+      "loss": 1.4848,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0352142554094188,
+      "grad_norm": 0.15762414421336599,
+      "learning_rate": 4.827223033142706e-05,
+      "loss": 1.4567,
+      "step": 1220
+    },
+    {
+      "epoch": 1.0436996181586762,
+      "grad_norm": 0.15058229398130366,
+      "learning_rate": 4.8217698735311414e-05,
+      "loss": 1.4672,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0521849809079338,
+      "grad_norm": 0.16010992835678386,
+      "learning_rate": 4.8162351680370044e-05,
+      "loss": 1.4458,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0606703436571914,
+      "grad_norm": 0.16758816000341356,
+      "learning_rate": 4.810619111051847e-05,
+      "loss": 1.4842,
+      "step": 1250
+    },
+    {
+      "epoch": 1.069155706406449,
+      "grad_norm": 0.16559260972674986,
+      "learning_rate": 4.8049218998244696e-05,
+      "loss": 1.4556,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0776410691557063,
+      "grad_norm": 0.17237632034416966,
+      "learning_rate": 4.7991437344539966e-05,
+      "loss": 1.4813,
+      "step": 1270
+    },
+    {
+      "epoch": 1.086126431904964,
+      "grad_norm": 0.17112756741722487,
+      "learning_rate": 4.793284817882845e-05,
+      "loss": 1.4535,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0946117946542215,
+      "grad_norm": 0.16828572707718548,
+      "learning_rate": 4.787345355889604e-05,
+      "loss": 1.4344,
+      "step": 1290
+    },
+    {
+      "epoch": 1.103097157403479,
+      "grad_norm": 0.15709986047041227,
+      "learning_rate": 4.7813255570817985e-05,
+      "loss": 1.4744,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1115825201527365,
+      "grad_norm": 0.16651547128146313,
+      "learning_rate": 4.775225632888568e-05,
+      "loss": 1.4561,
+      "step": 1310
+    },
+    {
+      "epoch": 1.120067882901994,
+      "grad_norm": 0.16750176017515714,
+      "learning_rate": 4.76904579755324e-05,
+      "loss": 1.4616,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1285532456512515,
+      "grad_norm": 0.1608016567554825,
+      "learning_rate": 4.7627862681258037e-05,
+      "loss": 1.4593,
+      "step": 1330
+    },
+    {
+      "epoch": 1.137038608400509,
+      "grad_norm": 0.21390766919038295,
+      "learning_rate": 4.756447264455287e-05,
+      "loss": 1.4484,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1455239711497667,
+      "grad_norm": 0.16826883293172662,
+      "learning_rate": 4.750029009182038e-05,
+      "loss": 1.4703,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1540093338990243,
+      "grad_norm": 0.17431508867079595,
+      "learning_rate": 4.7435317277299e-05,
+      "loss": 1.4701,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1624946966482816,
+      "grad_norm": 0.15973851467570443,
+      "learning_rate": 4.736955648298299e-05,
+      "loss": 1.4503,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1709800593975392,
+      "grad_norm": 0.1887713767970947,
+      "learning_rate": 4.730301001854225e-05,
+      "loss": 1.4624,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1794654221467968,
+      "grad_norm": 0.16898695344997974,
+      "learning_rate": 4.7235680221241216e-05,
+      "loss": 1.4452,
+      "step": 1390
+    },
+    {
+      "epoch": 1.1879507848960542,
+      "grad_norm": 0.20014553287073528,
+      "learning_rate": 4.716756945585681e-05,
+      "loss": 1.4717,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1964361476453118,
+      "grad_norm": 0.17137954325200072,
+      "learning_rate": 4.709868011459528e-05,
+      "loss": 1.4403,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2049215103945694,
+      "grad_norm": 0.17801721751888322,
+      "learning_rate": 4.7029014617008294e-05,
+      "loss": 1.4339,
+      "step": 1420
+    },
+    {
+      "epoch": 1.213406873143827,
+      "grad_norm": 0.17139613676642362,
+      "learning_rate": 4.695857540990789e-05,
+      "loss": 1.4573,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2218922358930844,
+      "grad_norm": 0.16971403514498054,
+      "learning_rate": 4.688736496728058e-05,
+      "loss": 1.4282,
+      "step": 1440
+    },
+    {
+      "epoch": 1.230377598642342,
+      "grad_norm": 0.17200272420880428,
+      "learning_rate": 4.681538579020038e-05,
+      "loss": 1.4434,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2388629613915996,
+      "grad_norm": 0.17208160407432616,
+      "learning_rate": 4.6742640406741106e-05,
+      "loss": 1.45,
+      "step": 1460
+    },
+    {
+      "epoch": 1.247348324140857,
+      "grad_norm": 0.1939626212901777,
+      "learning_rate": 4.666913137188743e-05,
+      "loss": 1.4608,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2558336868901145,
+      "grad_norm": 0.17291794493304186,
+      "learning_rate": 4.6594861267445236e-05,
+      "loss": 1.4671,
+      "step": 1480
+    },
+    {
+      "epoch": 1.2643190496393721,
+      "grad_norm": 0.18219792041638924,
+      "learning_rate": 4.651983270195093e-05,
+      "loss": 1.4262,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2728044123886297,
+      "grad_norm": 0.18086437830489926,
+      "learning_rate": 4.644404831057979e-05,
+      "loss": 1.4455,
+      "step": 1500
+    },
+    {
+      "epoch": 1.281289775137887,
+      "grad_norm": 0.17417619624549402,
+      "learning_rate": 4.636751075505344e-05,
+      "loss": 1.4873,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2897751378871447,
+      "grad_norm": 0.18354282411845188,
+      "learning_rate": 4.629022272354637e-05,
+      "loss": 1.4525,
+      "step": 1520
+    },
+    {
+      "epoch": 1.298260500636402,
+      "grad_norm": 0.17985617345325455,
+      "learning_rate": 4.621218693059149e-05,
+      "loss": 1.4303,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3067458633856597,
+      "grad_norm": 0.1809708317849863,
+      "learning_rate": 4.6133406116984795e-05,
+      "loss": 1.4631,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3152312261349173,
+      "grad_norm": 0.17487374671212322,
+      "learning_rate": 4.6053883049689145e-05,
+      "loss": 1.4482,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3237165888841749,
+      "grad_norm": 0.19912807671077193,
+      "learning_rate": 4.5973620521737036e-05,
+      "loss": 1.4497,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3322019516334322,
+      "grad_norm": 0.17853627546912074,
+      "learning_rate": 4.5892621352132514e-05,
+      "loss": 1.4456,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3406873143826898,
+      "grad_norm": 0.18252596927754394,
+      "learning_rate": 4.581088838575218e-05,
+      "loss": 1.4328,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3491726771319474,
+      "grad_norm": 0.17604951053556211,
+      "learning_rate": 4.572842449324525e-05,
+      "loss": 1.4442,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3576580398812048,
+      "grad_norm": 0.18358942463311748,
+      "learning_rate": 4.564523257093275e-05,
+      "loss": 1.4338,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3661434026304624,
+      "grad_norm": 0.20508703236267142,
+      "learning_rate": 4.5561315540705774e-05,
+      "loss": 1.4445,
+      "step": 1610
+    },
+    {
+      "epoch": 1.37462876537972,
+      "grad_norm": 0.18486352550747187,
+      "learning_rate": 4.547667634992288e-05,
+      "loss": 1.4261,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3831141281289776,
+      "grad_norm": 0.17492766465456316,
+      "learning_rate": 4.539131797130656e-05,
+      "loss": 1.4258,
+      "step": 1630
+    },
+    {
+      "epoch": 1.391599490878235,
+      "grad_norm": 0.19692876587833674,
+      "learning_rate": 4.530524340283881e-05,
+      "loss": 1.4349,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4000848536274926,
+      "grad_norm": 0.19155373430892478,
+      "learning_rate": 4.521845566765589e-05,
+      "loss": 1.4536,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4085702163767502,
+      "grad_norm": 0.18544325977459192,
+      "learning_rate": 4.513095781394208e-05,
+      "loss": 1.4363,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4170555791260075,
+      "grad_norm": 0.177828004720666,
+      "learning_rate": 4.504275291482267e-05,
+      "loss": 1.4595,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4255409418752651,
+      "grad_norm": 0.17855432230356816,
+      "learning_rate": 4.495384406825601e-05,
+      "loss": 1.4211,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4340263046245227,
+      "grad_norm": 0.20232492538380317,
+      "learning_rate": 4.486423439692469e-05,
+      "loss": 1.4189,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4425116673737803,
+      "grad_norm": 0.1975109303350431,
+      "learning_rate": 4.477392704812585e-05,
+      "loss": 1.4565,
+      "step": 1700
+    },
+    {
+      "epoch": 1.4509970301230377,
+      "grad_norm": 0.19619010830399825,
+      "learning_rate": 4.468292519366071e-05,
+      "loss": 1.4382,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4594823928722953,
+      "grad_norm": 0.18168826428246143,
+      "learning_rate": 4.459123202972308e-05,
+      "loss": 1.4471,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4679677556215527,
+      "grad_norm": 0.1923264062362399,
+      "learning_rate": 4.449885077678717e-05,
+      "loss": 1.4153,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4764531183708103,
+      "grad_norm": 0.1907937313040222,
+      "learning_rate": 4.440578467949445e-05,
+      "loss": 1.4432,
+      "step": 1740
+    },
+    {
+      "epoch": 1.4849384811200679,
+      "grad_norm": 0.19107457667767244,
+      "learning_rate": 4.431203700653968e-05,
+      "loss": 1.4285,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4934238438693255,
+      "grad_norm": 0.19847350429107552,
+      "learning_rate": 4.421761105055613e-05,
+      "loss": 1.4383,
+      "step": 1760
+    },
+    {
+      "epoch": 1.501909206618583,
+      "grad_norm": 0.18536475556610216,
+      "learning_rate": 4.4122510127999937e-05,
+      "loss": 1.42,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5103945693678404,
+      "grad_norm": 0.18481023473586697,
+      "learning_rate": 4.4026737579033584e-05,
+      "loss": 1.4384,
+      "step": 1780
+    },
+    {
+      "epoch": 1.518879932117098,
+      "grad_norm": 0.20863867505874642,
+      "learning_rate": 4.393029676740864e-05,
+      "loss": 1.4543,
+      "step": 1790
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "grad_norm": 0.1816036870853105,
+      "learning_rate": 4.3833191080347575e-05,
+      "loss": 1.434,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "eval_loss": 1.4622184038162231,
+      "eval_runtime": 52.4041,
+      "eval_samples_per_second": 7.27,
+      "eval_steps_per_second": 0.916,
+      "step": 1800
+    },
+    {
+      "epoch": 1.535850657615613,
+      "grad_norm": 0.19378252368958881,
+      "learning_rate": 4.3735423928424815e-05,
+      "loss": 1.4275,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5443360203648706,
+      "grad_norm": 0.20453331251433848,
+      "learning_rate": 4.363699874544697e-05,
+      "loss": 1.4203,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5528213831141282,
+      "grad_norm": 0.26684319417219377,
+      "learning_rate": 4.3537918988332156e-05,
+      "loss": 1.4372,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5613067458633858,
+      "grad_norm": 0.25745160303419773,
+      "learning_rate": 4.343818813698868e-05,
+      "loss": 1.4082,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5697921086126432,
+      "grad_norm": 0.19969727996700776,
+      "learning_rate": 4.3337809694192765e-05,
+      "loss": 1.4314,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5782774713619008,
+      "grad_norm": 0.20117210832277968,
+      "learning_rate": 4.3236787185465525e-05,
+      "loss": 1.4293,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5867628341111581,
+      "grad_norm": 0.20173003641028897,
+      "learning_rate": 4.313512415894913e-05,
+      "loss": 1.4406,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5952481968604157,
+      "grad_norm": 0.20304770794371527,
+      "learning_rate": 4.303282418528224e-05,
+      "loss": 1.4286,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6037335596096733,
+      "grad_norm": 0.19126658907738198,
+      "learning_rate": 4.292989085747452e-05,
+      "loss": 1.4184,
+      "step": 1890
+    },
+    {
+      "epoch": 1.612218922358931,
+      "grad_norm": 0.20069554966453027,
+      "learning_rate": 4.282632779078051e-05,
+      "loss": 1.4133,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6207042851081885,
+      "grad_norm": 0.1952881519566686,
+      "learning_rate": 4.2722138622572624e-05,
+      "loss": 1.4432,
+      "step": 1910
+    },
+    {
+      "epoch": 1.629189647857446,
+      "grad_norm": 0.19763704668680288,
+      "learning_rate": 4.261732701221339e-05,
+      "loss": 1.3921,
+      "step": 1920
+    },
+    {
+      "epoch": 1.6376750106067033,
+      "grad_norm": 0.19821464294464497,
+      "learning_rate": 4.2511896640926925e-05,
+      "loss": 1.4454,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6461603733559609,
+      "grad_norm": 0.20456545626297834,
+      "learning_rate": 4.240585121166966e-05,
+      "loss": 1.4147,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6546457361052185,
+      "grad_norm": 0.2119092529186395,
+      "learning_rate": 4.229919444900027e-05,
+      "loss": 1.3969,
+      "step": 1950
+    },
+    {
+      "epoch": 1.663131098854476,
+      "grad_norm": 0.20330157582122357,
+      "learning_rate": 4.2191930098948865e-05,
+      "loss": 1.426,
+      "step": 1960
+    },
+    {
+      "epoch": 1.6716164616037337,
+      "grad_norm": 0.21761164739298738,
+      "learning_rate": 4.2084061928885406e-05,
+      "loss": 1.4246,
+      "step": 1970
+    },
+    {
+      "epoch": 1.680101824352991,
+      "grad_norm": 0.19331588142071401,
+      "learning_rate": 4.197559372738741e-05,
+      "loss": 1.4305,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6885871871022486,
+      "grad_norm": 0.20188460724329996,
+      "learning_rate": 4.186652930410685e-05,
+      "loss": 1.4153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.697072549851506,
+      "grad_norm": 0.20988950033571588,
+      "learning_rate": 4.1756872489636425e-05,
+      "loss": 1.3894,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5890,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1681059326459904.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2000/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

checkpoint-2500/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-72B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.12.0

checkpoint-2500/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-2500/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-2500/latest

@@ -0,0 +1 @@
+global_step2500

checkpoint-2500/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-2500/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-2500/trainer_state.json

@@ -0,0 +1,1815 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1213406873143827,
+  "eval_steps": 600,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00848536274925753,
+      "grad_norm": 0.4898678891363344,
+      "learning_rate": 8.488964346349746e-07,
+      "loss": 1.8056,
+      "step": 10
+    },
+    {
+      "epoch": 0.01697072549851506,
+      "grad_norm": 0.3537473179717183,
+      "learning_rate": 1.6977928692699491e-06,
+      "loss": 1.7621,
+      "step": 20
+    },
+    {
+      "epoch": 0.025456088247772592,
+      "grad_norm": 0.28215953004159977,
+      "learning_rate": 2.546689303904924e-06,
+      "loss": 1.7571,
+      "step": 30
+    },
+    {
+      "epoch": 0.03394145099703012,
+      "grad_norm": 0.27446565146764923,
+      "learning_rate": 3.3955857385398982e-06,
+      "loss": 1.7136,
+      "step": 40
+    },
+    {
+      "epoch": 0.04242681374628765,
+      "grad_norm": 0.17051549768176558,
+      "learning_rate": 4.244482173174873e-06,
+      "loss": 1.6767,
+      "step": 50
+    },
+    {
+      "epoch": 0.050912176495545185,
+      "grad_norm": 0.17763882467320422,
+      "learning_rate": 5.093378607809848e-06,
+      "loss": 1.6371,
+      "step": 60
+    },
+    {
+      "epoch": 0.05939753924480271,
+      "grad_norm": 0.14311462596290048,
+      "learning_rate": 5.942275042444822e-06,
+      "loss": 1.6324,
+      "step": 70
+    },
+    {
+      "epoch": 0.06788290199406025,
+      "grad_norm": 0.1659540846071645,
+      "learning_rate": 6.7911714770797965e-06,
+      "loss": 1.6062,
+      "step": 80
+    },
+    {
+      "epoch": 0.07636826474331777,
+      "grad_norm": 0.20064072815620043,
+      "learning_rate": 7.640067911714771e-06,
+      "loss": 1.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0848536274925753,
+      "grad_norm": 0.2179045681711979,
+      "learning_rate": 8.488964346349745e-06,
+      "loss": 1.5898,
+      "step": 100
+    },
+    {
+      "epoch": 0.09333899024183284,
+      "grad_norm": 0.23866012053128668,
+      "learning_rate": 9.337860780984721e-06,
+      "loss": 1.5924,
+      "step": 110
+    },
+    {
+      "epoch": 0.10182435299109037,
+      "grad_norm": 0.18578051776430282,
+      "learning_rate": 1.0186757215619695e-05,
+      "loss": 1.5877,
+      "step": 120
+    },
+    {
+      "epoch": 0.1103097157403479,
+      "grad_norm": 0.2216509707409362,
+      "learning_rate": 1.103565365025467e-05,
+      "loss": 1.5947,
+      "step": 130
+    },
+    {
+      "epoch": 0.11879507848960542,
+      "grad_norm": 0.20427142255694086,
+      "learning_rate": 1.1884550084889643e-05,
+      "loss": 1.5841,
+      "step": 140
+    },
+    {
+      "epoch": 0.12728044123886295,
+      "grad_norm": 0.1765851415675038,
+      "learning_rate": 1.2733446519524619e-05,
+      "loss": 1.5878,
+      "step": 150
+    },
+    {
+      "epoch": 0.1357658039881205,
+      "grad_norm": 0.1769355117060811,
+      "learning_rate": 1.3582342954159593e-05,
+      "loss": 1.5795,
+      "step": 160
+    },
+    {
+      "epoch": 0.14425116673737803,
+      "grad_norm": 0.1617675663096666,
+      "learning_rate": 1.4431239388794569e-05,
+      "loss": 1.5549,
+      "step": 170
+    },
+    {
+      "epoch": 0.15273652948663555,
+      "grad_norm": 0.17302259072151574,
+      "learning_rate": 1.5280135823429543e-05,
+      "loss": 1.5808,
+      "step": 180
+    },
+    {
+      "epoch": 0.1612218922358931,
+      "grad_norm": 0.16876039012432806,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 1.5676,
+      "step": 190
+    },
+    {
+      "epoch": 0.1697072549851506,
+      "grad_norm": 0.19627360154037596,
+      "learning_rate": 1.697792869269949e-05,
+      "loss": 1.5598,
+      "step": 200
+    },
+    {
+      "epoch": 0.17819261773440814,
+      "grad_norm": 0.16078510362361015,
+      "learning_rate": 1.7826825127334465e-05,
+      "loss": 1.5667,
+      "step": 210
+    },
+    {
+      "epoch": 0.18667798048366568,
+      "grad_norm": 0.16044786518959703,
+      "learning_rate": 1.8675721561969442e-05,
+      "loss": 1.5815,
+      "step": 220
+    },
+    {
+      "epoch": 0.1951633432329232,
+      "grad_norm": 0.15656958873834717,
+      "learning_rate": 1.9524617996604416e-05,
+      "loss": 1.5576,
+      "step": 230
+    },
+    {
+      "epoch": 0.20364870598218074,
+      "grad_norm": 0.1687290471357602,
+      "learning_rate": 2.037351443123939e-05,
+      "loss": 1.5453,
+      "step": 240
+    },
+    {
+      "epoch": 0.21213406873143828,
+      "grad_norm": 0.1519017348276184,
+      "learning_rate": 2.1222410865874364e-05,
+      "loss": 1.5554,
+      "step": 250
+    },
+    {
+      "epoch": 0.2206194314806958,
+      "grad_norm": 0.15761892005160086,
+      "learning_rate": 2.207130730050934e-05,
+      "loss": 1.5494,
+      "step": 260
+    },
+    {
+      "epoch": 0.22910479422995333,
+      "grad_norm": 0.16857088482977495,
+      "learning_rate": 2.2920203735144312e-05,
+      "loss": 1.5794,
+      "step": 270
+    },
+    {
+      "epoch": 0.23759015697921085,
+      "grad_norm": 0.1678705209913503,
+      "learning_rate": 2.3769100169779286e-05,
+      "loss": 1.5373,
+      "step": 280
+    },
+    {
+      "epoch": 0.2460755197284684,
+      "grad_norm": 0.14812649566587394,
+      "learning_rate": 2.461799660441426e-05,
+      "loss": 1.5504,
+      "step": 290
+    },
+    {
+      "epoch": 0.2545608824777259,
+      "grad_norm": 0.17651916734325857,
+      "learning_rate": 2.5466893039049238e-05,
+      "loss": 1.5607,
+      "step": 300
+    },
+    {
+      "epoch": 0.26304624522698344,
+      "grad_norm": 0.14883055338507856,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 1.5311,
+      "step": 310
+    },
+    {
+      "epoch": 0.271531607976241,
+      "grad_norm": 0.15787522753231265,
+      "learning_rate": 2.7164685908319186e-05,
+      "loss": 1.5656,
+      "step": 320
+    },
+    {
+      "epoch": 0.2800169707254985,
+      "grad_norm": 0.1625232940237689,
+      "learning_rate": 2.801358234295416e-05,
+      "loss": 1.5686,
+      "step": 330
+    },
+    {
+      "epoch": 0.28850233347475607,
+      "grad_norm": 0.18505951289343867,
+      "learning_rate": 2.8862478777589137e-05,
+      "loss": 1.5474,
+      "step": 340
+    },
+    {
+      "epoch": 0.29698769622401355,
+      "grad_norm": 0.13785772316349984,
+      "learning_rate": 2.9711375212224108e-05,
+      "loss": 1.5696,
+      "step": 350
+    },
+    {
+      "epoch": 0.3054730589732711,
+      "grad_norm": 0.13531274658248552,
+      "learning_rate": 3.0560271646859086e-05,
+      "loss": 1.5551,
+      "step": 360
+    },
+    {
+      "epoch": 0.31395842172252864,
+      "grad_norm": 0.1366381415368909,
+      "learning_rate": 3.140916808149406e-05,
+      "loss": 1.524,
+      "step": 370
+    },
+    {
+      "epoch": 0.3224437844717862,
+      "grad_norm": 0.14587220569353926,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 1.5515,
+      "step": 380
+    },
+    {
+      "epoch": 0.3309291472210437,
+      "grad_norm": 0.13336349383744864,
+      "learning_rate": 3.310696095076401e-05,
+      "loss": 1.5457,
+      "step": 390
+    },
+    {
+      "epoch": 0.3394145099703012,
+      "grad_norm": 0.1772016947970983,
+      "learning_rate": 3.395585738539898e-05,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 0.34789987271955874,
+      "grad_norm": 0.13819420575084573,
+      "learning_rate": 3.4804753820033956e-05,
+      "loss": 1.5326,
+      "step": 410
+    },
+    {
+      "epoch": 0.3563852354688163,
+      "grad_norm": 0.12729862167862188,
+      "learning_rate": 3.565365025466893e-05,
+      "loss": 1.5387,
+      "step": 420
+    },
+    {
+      "epoch": 0.3648705982180738,
+      "grad_norm": 0.11777082851399363,
+      "learning_rate": 3.6502546689303904e-05,
+      "loss": 1.5587,
+      "step": 430
+    },
+    {
+      "epoch": 0.37335596096733137,
+      "grad_norm": 0.15372268131323022,
+      "learning_rate": 3.7351443123938885e-05,
+      "loss": 1.5362,
+      "step": 440
+    },
+    {
+      "epoch": 0.3818413237165889,
+      "grad_norm": 0.12616185572252248,
+      "learning_rate": 3.820033955857386e-05,
+      "loss": 1.5548,
+      "step": 450
+    },
+    {
+      "epoch": 0.3903266864658464,
+      "grad_norm": 0.1311200786303391,
+      "learning_rate": 3.904923599320883e-05,
+      "loss": 1.5409,
+      "step": 460
+    },
+    {
+      "epoch": 0.39881204921510394,
+      "grad_norm": 0.1707919112561785,
+      "learning_rate": 3.989813242784381e-05,
+      "loss": 1.5509,
+      "step": 470
+    },
+    {
+      "epoch": 0.4072974119643615,
+      "grad_norm": 0.14660149264284913,
+      "learning_rate": 4.074702886247878e-05,
+      "loss": 1.5433,
+      "step": 480
+    },
+    {
+      "epoch": 0.415782774713619,
+      "grad_norm": 0.12478895483834351,
+      "learning_rate": 4.1595925297113755e-05,
+      "loss": 1.5382,
+      "step": 490
+    },
+    {
+      "epoch": 0.42426813746287656,
+      "grad_norm": 0.12327957445795817,
+      "learning_rate": 4.244482173174873e-05,
+      "loss": 1.5515,
+      "step": 500
+    },
+    {
+      "epoch": 0.43275350021213405,
+      "grad_norm": 0.12922777738650987,
+      "learning_rate": 4.32937181663837e-05,
+      "loss": 1.5688,
+      "step": 510
+    },
+    {
+      "epoch": 0.4412388629613916,
+      "grad_norm": 0.12486802189783415,
+      "learning_rate": 4.414261460101868e-05,
+      "loss": 1.5452,
+      "step": 520
+    },
+    {
+      "epoch": 0.44972422571064913,
+      "grad_norm": 0.1360610874577123,
+      "learning_rate": 4.499151103565366e-05,
+      "loss": 1.5493,
+      "step": 530
+    },
+    {
+      "epoch": 0.45820958845990667,
+      "grad_norm": 0.1884897685356775,
+      "learning_rate": 4.5840407470288625e-05,
+      "loss": 1.5511,
+      "step": 540
+    },
+    {
+      "epoch": 0.4666949512091642,
+      "grad_norm": 0.12446302384809525,
+      "learning_rate": 4.6689303904923606e-05,
+      "loss": 1.5458,
+      "step": 550
+    },
+    {
+      "epoch": 0.4751803139584217,
+      "grad_norm": 0.13169591804768588,
+      "learning_rate": 4.753820033955857e-05,
+      "loss": 1.5569,
+      "step": 560
+    },
+    {
+      "epoch": 0.48366567670767924,
+      "grad_norm": 0.1343809247449631,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 1.5408,
+      "step": 570
+    },
+    {
+      "epoch": 0.4921510394569368,
+      "grad_norm": 0.14024589853602,
+      "learning_rate": 4.923599320882852e-05,
+      "loss": 1.5487,
+      "step": 580
+    },
+    {
+      "epoch": 0.5006364022061943,
+      "grad_norm": 0.16240429253875313,
+      "learning_rate": 4.999999560970061e-05,
+      "loss": 1.5488,
+      "step": 590
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "grad_norm": 0.12575424857894482,
+      "learning_rate": 4.999946877563971e-05,
+      "loss": 1.532,
+      "step": 600
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "eval_loss": 1.519254446029663,
+      "eval_runtime": 53.3242,
+      "eval_samples_per_second": 7.145,
+      "eval_steps_per_second": 0.9,
+      "step": 600
+    },
+    {
+      "epoch": 0.5176071277047094,
+      "grad_norm": 0.18688482756329736,
+      "learning_rate": 4.999806390290309e-05,
+      "loss": 1.5544,
+      "step": 610
+    },
+    {
+      "epoch": 0.5260924904539669,
+      "grad_norm": 0.12425469431830571,
+      "learning_rate": 4.999578104083307e-05,
+      "loss": 1.5443,
+      "step": 620
+    },
+    {
+      "epoch": 0.5345778532032245,
+      "grad_norm": 0.1299027485420099,
+      "learning_rate": 4.999262026960902e-05,
+      "loss": 1.5569,
+      "step": 630
+    },
+    {
+      "epoch": 0.543063215952482,
+      "grad_norm": 0.11441754852508934,
+      "learning_rate": 4.998858170024449e-05,
+      "loss": 1.5316,
+      "step": 640
+    },
+    {
+      "epoch": 0.5515485787017395,
+      "grad_norm": 0.14888547248976478,
+      "learning_rate": 4.998366547458326e-05,
+      "loss": 1.5177,
+      "step": 650
+    },
+    {
+      "epoch": 0.560033941450997,
+      "grad_norm": 0.14859292774768867,
+      "learning_rate": 4.997787176529449e-05,
+      "loss": 1.5394,
+      "step": 660
+    },
+    {
+      "epoch": 0.5685193042002545,
+      "grad_norm": 0.12499154376539734,
+      "learning_rate": 4.997120077586651e-05,
+      "loss": 1.5554,
+      "step": 670
+    },
+    {
+      "epoch": 0.5770046669495121,
+      "grad_norm": 0.1218974898058821,
+      "learning_rate": 4.9963652740599774e-05,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.5854900296987696,
+      "grad_norm": 0.1273110498715124,
+      "learning_rate": 4.995522792459859e-05,
+      "loss": 1.5349,
+      "step": 690
+    },
+    {
+      "epoch": 0.5939753924480271,
+      "grad_norm": 0.12115412881719101,
+      "learning_rate": 4.994592662376183e-05,
+      "loss": 1.5419,
+      "step": 700
+    },
+    {
+      "epoch": 0.6024607551972847,
+      "grad_norm": 0.14855096330233286,
+      "learning_rate": 4.99357491647725e-05,
+      "loss": 1.513,
+      "step": 710
+    },
+    {
+      "epoch": 0.6109461179465422,
+      "grad_norm": 0.11407988659327956,
+      "learning_rate": 4.992469590508628e-05,
+      "loss": 1.5243,
+      "step": 720
+    },
+    {
+      "epoch": 0.6194314806957998,
+      "grad_norm": 0.1197712643781127,
+      "learning_rate": 4.9912767232919035e-05,
+      "loss": 1.5177,
+      "step": 730
+    },
+    {
+      "epoch": 0.6279168434450573,
+      "grad_norm": 0.12400515877262065,
+      "learning_rate": 4.9899963567233074e-05,
+      "loss": 1.5619,
+      "step": 740
+    },
+    {
+      "epoch": 0.6364022061943148,
+      "grad_norm": 0.12250385257708406,
+      "learning_rate": 4.988628535772249e-05,
+      "loss": 1.539,
+      "step": 750
+    },
+    {
+      "epoch": 0.6448875689435724,
+      "grad_norm": 0.1262441090496857,
+      "learning_rate": 4.987173308479738e-05,
+      "loss": 1.5195,
+      "step": 760
+    },
+    {
+      "epoch": 0.6533729316928298,
+      "grad_norm": 0.12459694416473029,
+      "learning_rate": 4.985630725956694e-05,
+      "loss": 1.5462,
+      "step": 770
+    },
+    {
+      "epoch": 0.6618582944420874,
+      "grad_norm": 0.12985189006106762,
+      "learning_rate": 4.9840008423821527e-05,
+      "loss": 1.5113,
+      "step": 780
+    },
+    {
+      "epoch": 0.6703436571913449,
+      "grad_norm": 0.12689306141471304,
+      "learning_rate": 4.9822837150013636e-05,
+      "loss": 1.5201,
+      "step": 790
+    },
+    {
+      "epoch": 0.6788290199406024,
+      "grad_norm": 0.15393156370587963,
+      "learning_rate": 4.980479404123778e-05,
+      "loss": 1.5121,
+      "step": 800
+    },
+    {
+      "epoch": 0.68731438268986,
+      "grad_norm": 0.13213701895207608,
+      "learning_rate": 4.978587973120931e-05,
+      "loss": 1.5307,
+      "step": 810
+    },
+    {
+      "epoch": 0.6957997454391175,
+      "grad_norm": 0.11561354931316294,
+      "learning_rate": 4.9766094884242184e-05,
+      "loss": 1.5316,
+      "step": 820
+    },
+    {
+      "epoch": 0.7042851081883751,
+      "grad_norm": 0.12414772399330044,
+      "learning_rate": 4.974544019522559e-05,
+      "loss": 1.5148,
+      "step": 830
+    },
+    {
+      "epoch": 0.7127704709376326,
+      "grad_norm": 0.1171652849153521,
+      "learning_rate": 4.972391638959959e-05,
+      "loss": 1.5096,
+      "step": 840
+    },
+    {
+      "epoch": 0.7212558336868902,
+      "grad_norm": 0.12868937349582316,
+      "learning_rate": 4.9701524223329585e-05,
+      "loss": 1.5282,
+      "step": 850
+    },
+    {
+      "epoch": 0.7297411964361477,
+      "grad_norm": 0.1200015077117309,
+      "learning_rate": 4.967826448287981e-05,
+      "loss": 1.5512,
+      "step": 860
+    },
+    {
+      "epoch": 0.7382265591854051,
+      "grad_norm": 0.12340885660045105,
+      "learning_rate": 4.96541379851857e-05,
+      "loss": 1.5394,
+      "step": 870
+    },
+    {
+      "epoch": 0.7467119219346627,
+      "grad_norm": 0.12976937691467555,
+      "learning_rate": 4.962914557762517e-05,
+      "loss": 1.51,
+      "step": 880
+    },
+    {
+      "epoch": 0.7551972846839202,
+      "grad_norm": 0.11912878476038466,
+      "learning_rate": 4.9603288137988905e-05,
+      "loss": 1.5294,
+      "step": 890
+    },
+    {
+      "epoch": 0.7636826474331778,
+      "grad_norm": 0.1299625480337927,
+      "learning_rate": 4.957656657444947e-05,
+      "loss": 1.507,
+      "step": 900
+    },
+    {
+      "epoch": 0.7721680101824353,
+      "grad_norm": 0.12380144459698468,
+      "learning_rate": 4.954898182552946e-05,
+      "loss": 1.5376,
+      "step": 910
+    },
+    {
+      "epoch": 0.7806533729316928,
+      "grad_norm": 0.13139339643682763,
+      "learning_rate": 4.9520534860068535e-05,
+      "loss": 1.5291,
+      "step": 920
+    },
+    {
+      "epoch": 0.7891387356809504,
+      "grad_norm": 0.13088956203983898,
+      "learning_rate": 4.949122667718935e-05,
+      "loss": 1.5239,
+      "step": 930
+    },
+    {
+      "epoch": 0.7976240984302079,
+      "grad_norm": 0.12586052988453703,
+      "learning_rate": 4.94610583062625e-05,
+      "loss": 1.5525,
+      "step": 940
+    },
+    {
+      "epoch": 0.8061094611794655,
+      "grad_norm": 0.12020996031652877,
+      "learning_rate": 4.943003080687035e-05,
+      "loss": 1.5525,
+      "step": 950
+    },
+    {
+      "epoch": 0.814594823928723,
+      "grad_norm": 0.12866375954060869,
+      "learning_rate": 4.9398145268769856e-05,
+      "loss": 1.5266,
+      "step": 960
+    },
+    {
+      "epoch": 0.8230801866779804,
+      "grad_norm": 0.13166136756817035,
+      "learning_rate": 4.936540281185423e-05,
+      "loss": 1.5041,
+      "step": 970
+    },
+    {
+      "epoch": 0.831565549427238,
+      "grad_norm": 0.12481946698483787,
+      "learning_rate": 4.933180458611364e-05,
+      "loss": 1.5271,
+      "step": 980
+    },
+    {
+      "epoch": 0.8400509121764955,
+      "grad_norm": 0.12264463761209114,
+      "learning_rate": 4.9297351771594844e-05,
+      "loss": 1.5354,
+      "step": 990
+    },
+    {
+      "epoch": 0.8485362749257531,
+      "grad_norm": 0.11985452856537594,
+      "learning_rate": 4.926204557835968e-05,
+      "loss": 1.5167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8570216376750106,
+      "grad_norm": 0.13125396521190327,
+      "learning_rate": 4.9225887246442634e-05,
+      "loss": 1.5282,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8655070004242681,
+      "grad_norm": 0.12730192328072554,
+      "learning_rate": 4.918887804580725e-05,
+      "loss": 1.5089,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8739923631735257,
+      "grad_norm": 0.12724644219344786,
+      "learning_rate": 4.915101927630153e-05,
+      "loss": 1.4964,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8824777259227832,
+      "grad_norm": 0.13578611501833232,
+      "learning_rate": 4.911231226761227e-05,
+      "loss": 1.5189,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8909630886720408,
+      "grad_norm": 0.13577513964986457,
+      "learning_rate": 4.90727583792184e-05,
+      "loss": 1.5149,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8994484514212983,
+      "grad_norm": 0.1269735011676505,
+      "learning_rate": 4.903235900034317e-05,
+      "loss": 1.5066,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9079338141705557,
+      "grad_norm": 0.13250058214235566,
+      "learning_rate": 4.899111554990543e-05,
+      "loss": 1.5129,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9164191769198133,
+      "grad_norm": 0.13130735246433495,
+      "learning_rate": 4.894902947646975e-05,
+      "loss": 1.5156,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9249045396690708,
+      "grad_norm": 0.1273580180253049,
+      "learning_rate": 4.890610225819553e-05,
+      "loss": 1.5324,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9333899024183284,
+      "grad_norm": 0.13155314243939242,
+      "learning_rate": 4.8862335402785136e-05,
+      "loss": 1.5106,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9418752651675859,
+      "grad_norm": 0.13564895211984299,
+      "learning_rate": 4.88177304474309e-05,
+      "loss": 1.5067,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9503606279168434,
+      "grad_norm": 0.12774735587114736,
+      "learning_rate": 4.877228895876115e-05,
+      "loss": 1.5182,
+      "step": 1120
+    },
+    {
+      "epoch": 0.958845990666101,
+      "grad_norm": 0.1307997709537685,
+      "learning_rate": 4.872601253278517e-05,
+      "loss": 1.4969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9673313534153585,
+      "grad_norm": 0.1304794845040634,
+      "learning_rate": 4.867890279483717e-05,
+      "loss": 1.5264,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9758167161646161,
+      "grad_norm": 0.13666141796489684,
+      "learning_rate": 4.8630961399519206e-05,
+      "loss": 1.5467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9843020789138736,
+      "grad_norm": 0.1370278303190263,
+      "learning_rate": 4.8582190030643e-05,
+      "loss": 1.5127,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9927874416631312,
+      "grad_norm": 0.1390936629299565,
+      "learning_rate": 4.8532590401170894e-05,
+      "loss": 1.5058,
+      "step": 1170
+    },
+    {
+      "epoch": 1.0012728044123886,
+      "grad_norm": 0.12934475548108287,
+      "learning_rate": 4.848216425315561e-05,
+      "loss": 1.5202,
+      "step": 1180
+    },
+    {
+      "epoch": 1.0097581671616462,
+      "grad_norm": 0.13898591683370803,
+      "learning_rate": 4.843091335767913e-05,
+      "loss": 1.4563,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "grad_norm": 0.17488231535826249,
+      "learning_rate": 4.837883951479043e-05,
+      "loss": 1.4402,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "eval_loss": 1.4955657720565796,
+      "eval_runtime": 52.424,
+      "eval_samples_per_second": 7.268,
+      "eval_steps_per_second": 0.916,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0267288926601612,
+      "grad_norm": 0.1536036344095855,
+      "learning_rate": 4.832594455344229e-05,
+      "loss": 1.4848,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0352142554094188,
+      "grad_norm": 0.15762414421336599,
+      "learning_rate": 4.827223033142706e-05,
+      "loss": 1.4567,
+      "step": 1220
+    },
+    {
+      "epoch": 1.0436996181586762,
+      "grad_norm": 0.15058229398130366,
+      "learning_rate": 4.8217698735311414e-05,
+      "loss": 1.4672,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0521849809079338,
+      "grad_norm": 0.16010992835678386,
+      "learning_rate": 4.8162351680370044e-05,
+      "loss": 1.4458,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0606703436571914,
+      "grad_norm": 0.16758816000341356,
+      "learning_rate": 4.810619111051847e-05,
+      "loss": 1.4842,
+      "step": 1250
+    },
+    {
+      "epoch": 1.069155706406449,
+      "grad_norm": 0.16559260972674986,
+      "learning_rate": 4.8049218998244696e-05,
+      "loss": 1.4556,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0776410691557063,
+      "grad_norm": 0.17237632034416966,
+      "learning_rate": 4.7991437344539966e-05,
+      "loss": 1.4813,
+      "step": 1270
+    },
+    {
+      "epoch": 1.086126431904964,
+      "grad_norm": 0.17112756741722487,
+      "learning_rate": 4.793284817882845e-05,
+      "loss": 1.4535,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0946117946542215,
+      "grad_norm": 0.16828572707718548,
+      "learning_rate": 4.787345355889604e-05,
+      "loss": 1.4344,
+      "step": 1290
+    },
+    {
+      "epoch": 1.103097157403479,
+      "grad_norm": 0.15709986047041227,
+      "learning_rate": 4.7813255570817985e-05,
+      "loss": 1.4744,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1115825201527365,
+      "grad_norm": 0.16651547128146313,
+      "learning_rate": 4.775225632888568e-05,
+      "loss": 1.4561,
+      "step": 1310
+    },
+    {
+      "epoch": 1.120067882901994,
+      "grad_norm": 0.16750176017515714,
+      "learning_rate": 4.76904579755324e-05,
+      "loss": 1.4616,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1285532456512515,
+      "grad_norm": 0.1608016567554825,
+      "learning_rate": 4.7627862681258037e-05,
+      "loss": 1.4593,
+      "step": 1330
+    },
+    {
+      "epoch": 1.137038608400509,
+      "grad_norm": 0.21390766919038295,
+      "learning_rate": 4.756447264455287e-05,
+      "loss": 1.4484,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1455239711497667,
+      "grad_norm": 0.16826883293172662,
+      "learning_rate": 4.750029009182038e-05,
+      "loss": 1.4703,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1540093338990243,
+      "grad_norm": 0.17431508867079595,
+      "learning_rate": 4.7435317277299e-05,
+      "loss": 1.4701,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1624946966482816,
+      "grad_norm": 0.15973851467570443,
+      "learning_rate": 4.736955648298299e-05,
+      "loss": 1.4503,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1709800593975392,
+      "grad_norm": 0.1887713767970947,
+      "learning_rate": 4.730301001854225e-05,
+      "loss": 1.4624,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1794654221467968,
+      "grad_norm": 0.16898695344997974,
+      "learning_rate": 4.7235680221241216e-05,
+      "loss": 1.4452,
+      "step": 1390
+    },
+    {
+      "epoch": 1.1879507848960542,
+      "grad_norm": 0.20014553287073528,
+      "learning_rate": 4.716756945585681e-05,
+      "loss": 1.4717,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1964361476453118,
+      "grad_norm": 0.17137954325200072,
+      "learning_rate": 4.709868011459528e-05,
+      "loss": 1.4403,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2049215103945694,
+      "grad_norm": 0.17801721751888322,
+      "learning_rate": 4.7029014617008294e-05,
+      "loss": 1.4339,
+      "step": 1420
+    },
+    {
+      "epoch": 1.213406873143827,
+      "grad_norm": 0.17139613676642362,
+      "learning_rate": 4.695857540990789e-05,
+      "loss": 1.4573,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2218922358930844,
+      "grad_norm": 0.16971403514498054,
+      "learning_rate": 4.688736496728058e-05,
+      "loss": 1.4282,
+      "step": 1440
+    },
+    {
+      "epoch": 1.230377598642342,
+      "grad_norm": 0.17200272420880428,
+      "learning_rate": 4.681538579020038e-05,
+      "loss": 1.4434,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2388629613915996,
+      "grad_norm": 0.17208160407432616,
+      "learning_rate": 4.6742640406741106e-05,
+      "loss": 1.45,
+      "step": 1460
+    },
+    {
+      "epoch": 1.247348324140857,
+      "grad_norm": 0.1939626212901777,
+      "learning_rate": 4.666913137188743e-05,
+      "loss": 1.4608,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2558336868901145,
+      "grad_norm": 0.17291794493304186,
+      "learning_rate": 4.6594861267445236e-05,
+      "loss": 1.4671,
+      "step": 1480
+    },
+    {
+      "epoch": 1.2643190496393721,
+      "grad_norm": 0.18219792041638924,
+      "learning_rate": 4.651983270195093e-05,
+      "loss": 1.4262,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2728044123886297,
+      "grad_norm": 0.18086437830489926,
+      "learning_rate": 4.644404831057979e-05,
+      "loss": 1.4455,
+      "step": 1500
+    },
+    {
+      "epoch": 1.281289775137887,
+      "grad_norm": 0.17417619624549402,
+      "learning_rate": 4.636751075505344e-05,
+      "loss": 1.4873,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2897751378871447,
+      "grad_norm": 0.18354282411845188,
+      "learning_rate": 4.629022272354637e-05,
+      "loss": 1.4525,
+      "step": 1520
+    },
+    {
+      "epoch": 1.298260500636402,
+      "grad_norm": 0.17985617345325455,
+      "learning_rate": 4.621218693059149e-05,
+      "loss": 1.4303,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3067458633856597,
+      "grad_norm": 0.1809708317849863,
+      "learning_rate": 4.6133406116984795e-05,
+      "loss": 1.4631,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3152312261349173,
+      "grad_norm": 0.17487374671212322,
+      "learning_rate": 4.6053883049689145e-05,
+      "loss": 1.4482,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3237165888841749,
+      "grad_norm": 0.19912807671077193,
+      "learning_rate": 4.5973620521737036e-05,
+      "loss": 1.4497,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3322019516334322,
+      "grad_norm": 0.17853627546912074,
+      "learning_rate": 4.5892621352132514e-05,
+      "loss": 1.4456,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3406873143826898,
+      "grad_norm": 0.18252596927754394,
+      "learning_rate": 4.581088838575218e-05,
+      "loss": 1.4328,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3491726771319474,
+      "grad_norm": 0.17604951053556211,
+      "learning_rate": 4.572842449324525e-05,
+      "loss": 1.4442,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3576580398812048,
+      "grad_norm": 0.18358942463311748,
+      "learning_rate": 4.564523257093275e-05,
+      "loss": 1.4338,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3661434026304624,
+      "grad_norm": 0.20508703236267142,
+      "learning_rate": 4.5561315540705774e-05,
+      "loss": 1.4445,
+      "step": 1610
+    },
+    {
+      "epoch": 1.37462876537972,
+      "grad_norm": 0.18486352550747187,
+      "learning_rate": 4.547667634992288e-05,
+      "loss": 1.4261,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3831141281289776,
+      "grad_norm": 0.17492766465456316,
+      "learning_rate": 4.539131797130656e-05,
+      "loss": 1.4258,
+      "step": 1630
+    },
+    {
+      "epoch": 1.391599490878235,
+      "grad_norm": 0.19692876587833674,
+      "learning_rate": 4.530524340283881e-05,
+      "loss": 1.4349,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4000848536274926,
+      "grad_norm": 0.19155373430892478,
+      "learning_rate": 4.521845566765589e-05,
+      "loss": 1.4536,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4085702163767502,
+      "grad_norm": 0.18544325977459192,
+      "learning_rate": 4.513095781394208e-05,
+      "loss": 1.4363,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4170555791260075,
+      "grad_norm": 0.177828004720666,
+      "learning_rate": 4.504275291482267e-05,
+      "loss": 1.4595,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4255409418752651,
+      "grad_norm": 0.17855432230356816,
+      "learning_rate": 4.495384406825601e-05,
+      "loss": 1.4211,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4340263046245227,
+      "grad_norm": 0.20232492538380317,
+      "learning_rate": 4.486423439692469e-05,
+      "loss": 1.4189,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4425116673737803,
+      "grad_norm": 0.1975109303350431,
+      "learning_rate": 4.477392704812585e-05,
+      "loss": 1.4565,
+      "step": 1700
+    },
+    {
+      "epoch": 1.4509970301230377,
+      "grad_norm": 0.19619010830399825,
+      "learning_rate": 4.468292519366071e-05,
+      "loss": 1.4382,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4594823928722953,
+      "grad_norm": 0.18168826428246143,
+      "learning_rate": 4.459123202972308e-05,
+      "loss": 1.4471,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4679677556215527,
+      "grad_norm": 0.1923264062362399,
+      "learning_rate": 4.449885077678717e-05,
+      "loss": 1.4153,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4764531183708103,
+      "grad_norm": 0.1907937313040222,
+      "learning_rate": 4.440578467949445e-05,
+      "loss": 1.4432,
+      "step": 1740
+    },
+    {
+      "epoch": 1.4849384811200679,
+      "grad_norm": 0.19107457667767244,
+      "learning_rate": 4.431203700653968e-05,
+      "loss": 1.4285,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4934238438693255,
+      "grad_norm": 0.19847350429107552,
+      "learning_rate": 4.421761105055613e-05,
+      "loss": 1.4383,
+      "step": 1760
+    },
+    {
+      "epoch": 1.501909206618583,
+      "grad_norm": 0.18536475556610216,
+      "learning_rate": 4.4122510127999937e-05,
+      "loss": 1.42,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5103945693678404,
+      "grad_norm": 0.18481023473586697,
+      "learning_rate": 4.4026737579033584e-05,
+      "loss": 1.4384,
+      "step": 1780
+    },
+    {
+      "epoch": 1.518879932117098,
+      "grad_norm": 0.20863867505874642,
+      "learning_rate": 4.393029676740864e-05,
+      "loss": 1.4543,
+      "step": 1790
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "grad_norm": 0.1816036870853105,
+      "learning_rate": 4.3833191080347575e-05,
+      "loss": 1.434,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "eval_loss": 1.4622184038162231,
+      "eval_runtime": 52.4041,
+      "eval_samples_per_second": 7.27,
+      "eval_steps_per_second": 0.916,
+      "step": 1800
+    },
+    {
+      "epoch": 1.535850657615613,
+      "grad_norm": 0.19378252368958881,
+      "learning_rate": 4.3735423928424815e-05,
+      "loss": 1.4275,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5443360203648706,
+      "grad_norm": 0.20453331251433848,
+      "learning_rate": 4.363699874544697e-05,
+      "loss": 1.4203,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5528213831141282,
+      "grad_norm": 0.26684319417219377,
+      "learning_rate": 4.3537918988332156e-05,
+      "loss": 1.4372,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5613067458633858,
+      "grad_norm": 0.25745160303419773,
+      "learning_rate": 4.343818813698868e-05,
+      "loss": 1.4082,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5697921086126432,
+      "grad_norm": 0.19969727996700776,
+      "learning_rate": 4.3337809694192765e-05,
+      "loss": 1.4314,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5782774713619008,
+      "grad_norm": 0.20117210832277968,
+      "learning_rate": 4.3236787185465525e-05,
+      "loss": 1.4293,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5867628341111581,
+      "grad_norm": 0.20173003641028897,
+      "learning_rate": 4.313512415894913e-05,
+      "loss": 1.4406,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5952481968604157,
+      "grad_norm": 0.20304770794371527,
+      "learning_rate": 4.303282418528224e-05,
+      "loss": 1.4286,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6037335596096733,
+      "grad_norm": 0.19126658907738198,
+      "learning_rate": 4.292989085747452e-05,
+      "loss": 1.4184,
+      "step": 1890
+    },
+    {
+      "epoch": 1.612218922358931,
+      "grad_norm": 0.20069554966453027,
+      "learning_rate": 4.282632779078051e-05,
+      "loss": 1.4133,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6207042851081885,
+      "grad_norm": 0.1952881519566686,
+      "learning_rate": 4.2722138622572624e-05,
+      "loss": 1.4432,
+      "step": 1910
+    },
+    {
+      "epoch": 1.629189647857446,
+      "grad_norm": 0.19763704668680288,
+      "learning_rate": 4.261732701221339e-05,
+      "loss": 1.3921,
+      "step": 1920
+    },
+    {
+      "epoch": 1.6376750106067033,
+      "grad_norm": 0.19821464294464497,
+      "learning_rate": 4.2511896640926925e-05,
+      "loss": 1.4454,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6461603733559609,
+      "grad_norm": 0.20456545626297834,
+      "learning_rate": 4.240585121166966e-05,
+      "loss": 1.4147,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6546457361052185,
+      "grad_norm": 0.2119092529186395,
+      "learning_rate": 4.229919444900027e-05,
+      "loss": 1.3969,
+      "step": 1950
+    },
+    {
+      "epoch": 1.663131098854476,
+      "grad_norm": 0.20330157582122357,
+      "learning_rate": 4.2191930098948865e-05,
+      "loss": 1.426,
+      "step": 1960
+    },
+    {
+      "epoch": 1.6716164616037337,
+      "grad_norm": 0.21761164739298738,
+      "learning_rate": 4.2084061928885406e-05,
+      "loss": 1.4246,
+      "step": 1970
+    },
+    {
+      "epoch": 1.680101824352991,
+      "grad_norm": 0.19331588142071401,
+      "learning_rate": 4.197559372738741e-05,
+      "loss": 1.4305,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6885871871022486,
+      "grad_norm": 0.20188460724329996,
+      "learning_rate": 4.186652930410685e-05,
+      "loss": 1.4153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.697072549851506,
+      "grad_norm": 0.20988950033571588,
+      "learning_rate": 4.1756872489636425e-05,
+      "loss": 1.3894,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7055579126007636,
+      "grad_norm": 0.1966475893123187,
+      "learning_rate": 4.1646627135374916e-05,
+      "loss": 1.3962,
+      "step": 2010
+    },
+    {
+      "epoch": 1.7140432753500212,
+      "grad_norm": 0.20785207367991768,
+      "learning_rate": 4.1535797113392004e-05,
+      "loss": 1.4037,
+      "step": 2020
+    },
+    {
+      "epoch": 1.7225286380992788,
+      "grad_norm": 0.2029940281663133,
+      "learning_rate": 4.1424386316292224e-05,
+      "loss": 1.4011,
+      "step": 2030
+    },
+    {
+      "epoch": 1.7310140008485364,
+      "grad_norm": 0.2247844551379277,
+      "learning_rate": 4.131239865707829e-05,
+      "loss": 1.4084,
+      "step": 2040
+    },
+    {
+      "epoch": 1.7394993635977938,
+      "grad_norm": 0.20900441746105022,
+      "learning_rate": 4.11998380690136e-05,
+      "loss": 1.4235,
+      "step": 2050
+    },
+    {
+      "epoch": 1.7479847263470514,
+      "grad_norm": 0.20362408546889926,
+      "learning_rate": 4.108670850548416e-05,
+      "loss": 1.4204,
+      "step": 2060
+    },
+    {
+      "epoch": 1.7564700890963088,
+      "grad_norm": 0.22281567946240438,
+      "learning_rate": 4.097301393985968e-05,
+      "loss": 1.4023,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7649554518455663,
+      "grad_norm": 0.20867113178797225,
+      "learning_rate": 4.085875836535404e-05,
+      "loss": 1.3895,
+      "step": 2080
+    },
+    {
+      "epoch": 1.773440814594824,
+      "grad_norm": 0.22113231886160947,
+      "learning_rate": 4.0743945794885063e-05,
+      "loss": 1.3963,
+      "step": 2090
+    },
+    {
+      "epoch": 1.7819261773440815,
+      "grad_norm": 0.22334563577844263,
+      "learning_rate": 4.062858026093351e-05,
+      "loss": 1.3988,
+      "step": 2100
+    },
+    {
+      "epoch": 1.7904115400933391,
+      "grad_norm": 0.23218581668265403,
+      "learning_rate": 4.051266581540152e-05,
+      "loss": 1.4068,
+      "step": 2110
+    },
+    {
+      "epoch": 1.7988969028425965,
+      "grad_norm": 0.20295589384571033,
+      "learning_rate": 4.0396206529470234e-05,
+      "loss": 1.3883,
+      "step": 2120
+    },
+    {
+      "epoch": 1.8073822655918539,
+      "grad_norm": 0.22861611442392848,
+      "learning_rate": 4.027920649345687e-05,
+      "loss": 1.4043,
+      "step": 2130
+    },
+    {
+      "epoch": 1.8158676283411115,
+      "grad_norm": 0.2083012771089638,
+      "learning_rate": 4.0161669816671e-05,
+      "loss": 1.398,
+      "step": 2140
+    },
+    {
+      "epoch": 1.824352991090369,
+      "grad_norm": 0.21936173231840464,
+      "learning_rate": 4.004360062727028e-05,
+      "loss": 1.4142,
+      "step": 2150
+    },
+    {
+      "epoch": 1.8328383538396267,
+      "grad_norm": 0.21383435796328337,
+      "learning_rate": 3.9925003072115406e-05,
+      "loss": 1.4138,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8413237165888843,
+      "grad_norm": 0.23301608248270392,
+      "learning_rate": 3.9805881316624506e-05,
+      "loss": 1.4195,
+      "step": 2170
+    },
+    {
+      "epoch": 1.8498090793381419,
+      "grad_norm": 0.22424766656883474,
+      "learning_rate": 3.968623954462681e-05,
+      "loss": 1.4011,
+      "step": 2180
+    },
+    {
+      "epoch": 1.8582944420873992,
+      "grad_norm": 0.21286417342881453,
+      "learning_rate": 3.9566081958215734e-05,
+      "loss": 1.409,
+      "step": 2190
+    },
+    {
+      "epoch": 1.8667798048366566,
+      "grad_norm": 0.21944800687444807,
+      "learning_rate": 3.9445412777601284e-05,
+      "loss": 1.3877,
+      "step": 2200
+    },
+    {
+      "epoch": 1.8752651675859142,
+      "grad_norm": 0.23113173625974803,
+      "learning_rate": 3.932423624096181e-05,
+      "loss": 1.4089,
+      "step": 2210
+    },
+    {
+      "epoch": 1.8837505303351718,
+      "grad_norm": 0.2081941699587778,
+      "learning_rate": 3.920255660429517e-05,
+      "loss": 1.4024,
+      "step": 2220
+    },
+    {
+      "epoch": 1.8922358930844294,
+      "grad_norm": 0.2188685806654701,
+      "learning_rate": 3.908037814126927e-05,
+      "loss": 1.3878,
+      "step": 2230
+    },
+    {
+      "epoch": 1.900721255833687,
+      "grad_norm": 0.22761843244757962,
+      "learning_rate": 3.895770514307193e-05,
+      "loss": 1.4004,
+      "step": 2240
+    },
+    {
+      "epoch": 1.9092066185829444,
+      "grad_norm": 0.23309183623120422,
+      "learning_rate": 3.883454191826017e-05,
+      "loss": 1.4188,
+      "step": 2250
+    },
+    {
+      "epoch": 1.917691981332202,
+      "grad_norm": 0.20329785843911802,
+      "learning_rate": 3.871089279260891e-05,
+      "loss": 1.3893,
+      "step": 2260
+    },
+    {
+      "epoch": 1.9261773440814594,
+      "grad_norm": 0.23470973193726366,
+      "learning_rate": 3.8586762108958995e-05,
+      "loss": 1.3974,
+      "step": 2270
+    },
+    {
+      "epoch": 1.934662706830717,
+      "grad_norm": 0.22779136837044714,
+      "learning_rate": 3.8462154227064725e-05,
+      "loss": 1.4115,
+      "step": 2280
+    },
+    {
+      "epoch": 1.9431480695799745,
+      "grad_norm": 0.22338952315651892,
+      "learning_rate": 3.833707352344068e-05,
+      "loss": 1.3873,
+      "step": 2290
+    },
+    {
+      "epoch": 1.9516334323292321,
+      "grad_norm": 0.23069304025882129,
+      "learning_rate": 3.821152439120801e-05,
+      "loss": 1.3944,
+      "step": 2300
+    },
+    {
+      "epoch": 1.9601187950784897,
+      "grad_norm": 0.23590596270163203,
+      "learning_rate": 3.808551123994018e-05,
+      "loss": 1.3857,
+      "step": 2310
+    },
+    {
+      "epoch": 1.9686041578277471,
+      "grad_norm": 0.22545661808214923,
+      "learning_rate": 3.795903849550805e-05,
+      "loss": 1.3628,
+      "step": 2320
+    },
+    {
+      "epoch": 1.9770895205770047,
+      "grad_norm": 0.2450769875954842,
+      "learning_rate": 3.7832110599924455e-05,
+      "loss": 1.4079,
+      "step": 2330
+    },
+    {
+      "epoch": 1.985574883326262,
+      "grad_norm": 0.22931499326784313,
+      "learning_rate": 3.7704732011188166e-05,
+      "loss": 1.379,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9940602460755197,
+      "grad_norm": 0.22417244507397657,
+      "learning_rate": 3.7576907203127346e-05,
+      "loss": 1.4035,
+      "step": 2350
+    },
+    {
+      "epoch": 2.0025456088247773,
+      "grad_norm": 0.24496197221575314,
+      "learning_rate": 3.7448640665242406e-05,
+      "loss": 1.442,
+      "step": 2360
+    },
+    {
+      "epoch": 2.011030971574035,
+      "grad_norm": 0.2532740296990078,
+      "learning_rate": 3.73199369025483e-05,
+      "loss": 1.2672,
+      "step": 2370
+    },
+    {
+      "epoch": 2.0195163343232925,
+      "grad_norm": 0.2890155987968593,
+      "learning_rate": 3.7190800435416355e-05,
+      "loss": 1.246,
+      "step": 2380
+    },
+    {
+      "epoch": 2.02800169707255,
+      "grad_norm": 0.2541972565696406,
+      "learning_rate": 3.706123579941545e-05,
+      "loss": 1.2603,
+      "step": 2390
+    },
+    {
+      "epoch": 2.036487059821807,
+      "grad_norm": 0.2530140862527023,
+      "learning_rate": 3.693124754515272e-05,
+      "loss": 1.2638,
+      "step": 2400
+    },
+    {
+      "epoch": 2.036487059821807,
+      "eval_loss": 1.435962438583374,
+      "eval_runtime": 52.582,
+      "eval_samples_per_second": 7.246,
+      "eval_steps_per_second": 0.913,
+      "step": 2400
+    },
+    {
+      "epoch": 2.044972422571065,
+      "grad_norm": 0.25100458343337734,
+      "learning_rate": 3.680084023811377e-05,
+      "loss": 1.2711,
+      "step": 2410
+    },
+    {
+      "epoch": 2.0534577853203224,
+      "grad_norm": 0.2695727673292618,
+      "learning_rate": 3.66700184585023e-05,
+      "loss": 1.2578,
+      "step": 2420
+    },
+    {
+      "epoch": 2.06194314806958,
+      "grad_norm": 0.2605068415443213,
+      "learning_rate": 3.6538786801079226e-05,
+      "loss": 1.2506,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0704285108188376,
+      "grad_norm": 0.27415607207865045,
+      "learning_rate": 3.64071498750013e-05,
+      "loss": 1.2852,
+      "step": 2440
+    },
+    {
+      "epoch": 2.078913873568095,
+      "grad_norm": 0.2688900338206285,
+      "learning_rate": 3.627511230365928e-05,
+      "loss": 1.2695,
+      "step": 2450
+    },
+    {
+      "epoch": 2.0873992363173524,
+      "grad_norm": 0.2750825805336503,
+      "learning_rate": 3.614267872451546e-05,
+      "loss": 1.2643,
+      "step": 2460
+    },
+    {
+      "epoch": 2.09588459906661,
+      "grad_norm": 0.2659269066581903,
+      "learning_rate": 3.600985378894086e-05,
+      "loss": 1.2868,
+      "step": 2470
+    },
+    {
+      "epoch": 2.1043699618158676,
+      "grad_norm": 0.24411151291321526,
+      "learning_rate": 3.587664216205183e-05,
+      "loss": 1.2571,
+      "step": 2480
+    },
+    {
+      "epoch": 2.112855324565125,
+      "grad_norm": 0.2574194755634052,
+      "learning_rate": 3.574304852254621e-05,
+      "loss": 1.2769,
+      "step": 2490
+    },
+    {
+      "epoch": 2.1213406873143827,
+      "grad_norm": 0.2894545074998905,
+      "learning_rate": 3.5609077562538997e-05,
+      "loss": 1.2469,
+      "step": 2500
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5890,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2101429250555904.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2500/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

checkpoint-3000/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-72B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.12.0

checkpoint-4000/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-4000/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-4000/latest

@@ -0,0 +1 @@
+global_step4000

checkpoint-4000/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-4000/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-4000/trainer_state.json

@@ -0,0 +1,2881 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.3941450997030125,
+  "eval_steps": 600,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00848536274925753,
+      "grad_norm": 0.4898678891363344,
+      "learning_rate": 8.488964346349746e-07,
+      "loss": 1.8056,
+      "step": 10
+    },
+    {
+      "epoch": 0.01697072549851506,
+      "grad_norm": 0.3537473179717183,
+      "learning_rate": 1.6977928692699491e-06,
+      "loss": 1.7621,
+      "step": 20
+    },
+    {
+      "epoch": 0.025456088247772592,
+      "grad_norm": 0.28215953004159977,
+      "learning_rate": 2.546689303904924e-06,
+      "loss": 1.7571,
+      "step": 30
+    },
+    {
+      "epoch": 0.03394145099703012,
+      "grad_norm": 0.27446565146764923,
+      "learning_rate": 3.3955857385398982e-06,
+      "loss": 1.7136,
+      "step": 40
+    },
+    {
+      "epoch": 0.04242681374628765,
+      "grad_norm": 0.17051549768176558,
+      "learning_rate": 4.244482173174873e-06,
+      "loss": 1.6767,
+      "step": 50
+    },
+    {
+      "epoch": 0.050912176495545185,
+      "grad_norm": 0.17763882467320422,
+      "learning_rate": 5.093378607809848e-06,
+      "loss": 1.6371,
+      "step": 60
+    },
+    {
+      "epoch": 0.05939753924480271,
+      "grad_norm": 0.14311462596290048,
+      "learning_rate": 5.942275042444822e-06,
+      "loss": 1.6324,
+      "step": 70
+    },
+    {
+      "epoch": 0.06788290199406025,
+      "grad_norm": 0.1659540846071645,
+      "learning_rate": 6.7911714770797965e-06,
+      "loss": 1.6062,
+      "step": 80
+    },
+    {
+      "epoch": 0.07636826474331777,
+      "grad_norm": 0.20064072815620043,
+      "learning_rate": 7.640067911714771e-06,
+      "loss": 1.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0848536274925753,
+      "grad_norm": 0.2179045681711979,
+      "learning_rate": 8.488964346349745e-06,
+      "loss": 1.5898,
+      "step": 100
+    },
+    {
+      "epoch": 0.09333899024183284,
+      "grad_norm": 0.23866012053128668,
+      "learning_rate": 9.337860780984721e-06,
+      "loss": 1.5924,
+      "step": 110
+    },
+    {
+      "epoch": 0.10182435299109037,
+      "grad_norm": 0.18578051776430282,
+      "learning_rate": 1.0186757215619695e-05,
+      "loss": 1.5877,
+      "step": 120
+    },
+    {
+      "epoch": 0.1103097157403479,
+      "grad_norm": 0.2216509707409362,
+      "learning_rate": 1.103565365025467e-05,
+      "loss": 1.5947,
+      "step": 130
+    },
+    {
+      "epoch": 0.11879507848960542,
+      "grad_norm": 0.20427142255694086,
+      "learning_rate": 1.1884550084889643e-05,
+      "loss": 1.5841,
+      "step": 140
+    },
+    {
+      "epoch": 0.12728044123886295,
+      "grad_norm": 0.1765851415675038,
+      "learning_rate": 1.2733446519524619e-05,
+      "loss": 1.5878,
+      "step": 150
+    },
+    {
+      "epoch": 0.1357658039881205,
+      "grad_norm": 0.1769355117060811,
+      "learning_rate": 1.3582342954159593e-05,
+      "loss": 1.5795,
+      "step": 160
+    },
+    {
+      "epoch": 0.14425116673737803,
+      "grad_norm": 0.1617675663096666,
+      "learning_rate": 1.4431239388794569e-05,
+      "loss": 1.5549,
+      "step": 170
+    },
+    {
+      "epoch": 0.15273652948663555,
+      "grad_norm": 0.17302259072151574,
+      "learning_rate": 1.5280135823429543e-05,
+      "loss": 1.5808,
+      "step": 180
+    },
+    {
+      "epoch": 0.1612218922358931,
+      "grad_norm": 0.16876039012432806,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 1.5676,
+      "step": 190
+    },
+    {
+      "epoch": 0.1697072549851506,
+      "grad_norm": 0.19627360154037596,
+      "learning_rate": 1.697792869269949e-05,
+      "loss": 1.5598,
+      "step": 200
+    },
+    {
+      "epoch": 0.17819261773440814,
+      "grad_norm": 0.16078510362361015,
+      "learning_rate": 1.7826825127334465e-05,
+      "loss": 1.5667,
+      "step": 210
+    },
+    {
+      "epoch": 0.18667798048366568,
+      "grad_norm": 0.16044786518959703,
+      "learning_rate": 1.8675721561969442e-05,
+      "loss": 1.5815,
+      "step": 220
+    },
+    {
+      "epoch": 0.1951633432329232,
+      "grad_norm": 0.15656958873834717,
+      "learning_rate": 1.9524617996604416e-05,
+      "loss": 1.5576,
+      "step": 230
+    },
+    {
+      "epoch": 0.20364870598218074,
+      "grad_norm": 0.1687290471357602,
+      "learning_rate": 2.037351443123939e-05,
+      "loss": 1.5453,
+      "step": 240
+    },
+    {
+      "epoch": 0.21213406873143828,
+      "grad_norm": 0.1519017348276184,
+      "learning_rate": 2.1222410865874364e-05,
+      "loss": 1.5554,
+      "step": 250
+    },
+    {
+      "epoch": 0.2206194314806958,
+      "grad_norm": 0.15761892005160086,
+      "learning_rate": 2.207130730050934e-05,
+      "loss": 1.5494,
+      "step": 260
+    },
+    {
+      "epoch": 0.22910479422995333,
+      "grad_norm": 0.16857088482977495,
+      "learning_rate": 2.2920203735144312e-05,
+      "loss": 1.5794,
+      "step": 270
+    },
+    {
+      "epoch": 0.23759015697921085,
+      "grad_norm": 0.1678705209913503,
+      "learning_rate": 2.3769100169779286e-05,
+      "loss": 1.5373,
+      "step": 280
+    },
+    {
+      "epoch": 0.2460755197284684,
+      "grad_norm": 0.14812649566587394,
+      "learning_rate": 2.461799660441426e-05,
+      "loss": 1.5504,
+      "step": 290
+    },
+    {
+      "epoch": 0.2545608824777259,
+      "grad_norm": 0.17651916734325857,
+      "learning_rate": 2.5466893039049238e-05,
+      "loss": 1.5607,
+      "step": 300
+    },
+    {
+      "epoch": 0.26304624522698344,
+      "grad_norm": 0.14883055338507856,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 1.5311,
+      "step": 310
+    },
+    {
+      "epoch": 0.271531607976241,
+      "grad_norm": 0.15787522753231265,
+      "learning_rate": 2.7164685908319186e-05,
+      "loss": 1.5656,
+      "step": 320
+    },
+    {
+      "epoch": 0.2800169707254985,
+      "grad_norm": 0.1625232940237689,
+      "learning_rate": 2.801358234295416e-05,
+      "loss": 1.5686,
+      "step": 330
+    },
+    {
+      "epoch": 0.28850233347475607,
+      "grad_norm": 0.18505951289343867,
+      "learning_rate": 2.8862478777589137e-05,
+      "loss": 1.5474,
+      "step": 340
+    },
+    {
+      "epoch": 0.29698769622401355,
+      "grad_norm": 0.13785772316349984,
+      "learning_rate": 2.9711375212224108e-05,
+      "loss": 1.5696,
+      "step": 350
+    },
+    {
+      "epoch": 0.3054730589732711,
+      "grad_norm": 0.13531274658248552,
+      "learning_rate": 3.0560271646859086e-05,
+      "loss": 1.5551,
+      "step": 360
+    },
+    {
+      "epoch": 0.31395842172252864,
+      "grad_norm": 0.1366381415368909,
+      "learning_rate": 3.140916808149406e-05,
+      "loss": 1.524,
+      "step": 370
+    },
+    {
+      "epoch": 0.3224437844717862,
+      "grad_norm": 0.14587220569353926,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 1.5515,
+      "step": 380
+    },
+    {
+      "epoch": 0.3309291472210437,
+      "grad_norm": 0.13336349383744864,
+      "learning_rate": 3.310696095076401e-05,
+      "loss": 1.5457,
+      "step": 390
+    },
+    {
+      "epoch": 0.3394145099703012,
+      "grad_norm": 0.1772016947970983,
+      "learning_rate": 3.395585738539898e-05,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 0.34789987271955874,
+      "grad_norm": 0.13819420575084573,
+      "learning_rate": 3.4804753820033956e-05,
+      "loss": 1.5326,
+      "step": 410
+    },
+    {
+      "epoch": 0.3563852354688163,
+      "grad_norm": 0.12729862167862188,
+      "learning_rate": 3.565365025466893e-05,
+      "loss": 1.5387,
+      "step": 420
+    },
+    {
+      "epoch": 0.3648705982180738,
+      "grad_norm": 0.11777082851399363,
+      "learning_rate": 3.6502546689303904e-05,
+      "loss": 1.5587,
+      "step": 430
+    },
+    {
+      "epoch": 0.37335596096733137,
+      "grad_norm": 0.15372268131323022,
+      "learning_rate": 3.7351443123938885e-05,
+      "loss": 1.5362,
+      "step": 440
+    },
+    {
+      "epoch": 0.3818413237165889,
+      "grad_norm": 0.12616185572252248,
+      "learning_rate": 3.820033955857386e-05,
+      "loss": 1.5548,
+      "step": 450
+    },
+    {
+      "epoch": 0.3903266864658464,
+      "grad_norm": 0.1311200786303391,
+      "learning_rate": 3.904923599320883e-05,
+      "loss": 1.5409,
+      "step": 460
+    },
+    {
+      "epoch": 0.39881204921510394,
+      "grad_norm": 0.1707919112561785,
+      "learning_rate": 3.989813242784381e-05,
+      "loss": 1.5509,
+      "step": 470
+    },
+    {
+      "epoch": 0.4072974119643615,
+      "grad_norm": 0.14660149264284913,
+      "learning_rate": 4.074702886247878e-05,
+      "loss": 1.5433,
+      "step": 480
+    },
+    {
+      "epoch": 0.415782774713619,
+      "grad_norm": 0.12478895483834351,
+      "learning_rate": 4.1595925297113755e-05,
+      "loss": 1.5382,
+      "step": 490
+    },
+    {
+      "epoch": 0.42426813746287656,
+      "grad_norm": 0.12327957445795817,
+      "learning_rate": 4.244482173174873e-05,
+      "loss": 1.5515,
+      "step": 500
+    },
+    {
+      "epoch": 0.43275350021213405,
+      "grad_norm": 0.12922777738650987,
+      "learning_rate": 4.32937181663837e-05,
+      "loss": 1.5688,
+      "step": 510
+    },
+    {
+      "epoch": 0.4412388629613916,
+      "grad_norm": 0.12486802189783415,
+      "learning_rate": 4.414261460101868e-05,
+      "loss": 1.5452,
+      "step": 520
+    },
+    {
+      "epoch": 0.44972422571064913,
+      "grad_norm": 0.1360610874577123,
+      "learning_rate": 4.499151103565366e-05,
+      "loss": 1.5493,
+      "step": 530
+    },
+    {
+      "epoch": 0.45820958845990667,
+      "grad_norm": 0.1884897685356775,
+      "learning_rate": 4.5840407470288625e-05,
+      "loss": 1.5511,
+      "step": 540
+    },
+    {
+      "epoch": 0.4666949512091642,
+      "grad_norm": 0.12446302384809525,
+      "learning_rate": 4.6689303904923606e-05,
+      "loss": 1.5458,
+      "step": 550
+    },
+    {
+      "epoch": 0.4751803139584217,
+      "grad_norm": 0.13169591804768588,
+      "learning_rate": 4.753820033955857e-05,
+      "loss": 1.5569,
+      "step": 560
+    },
+    {
+      "epoch": 0.48366567670767924,
+      "grad_norm": 0.1343809247449631,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 1.5408,
+      "step": 570
+    },
+    {
+      "epoch": 0.4921510394569368,
+      "grad_norm": 0.14024589853602,
+      "learning_rate": 4.923599320882852e-05,
+      "loss": 1.5487,
+      "step": 580
+    },
+    {
+      "epoch": 0.5006364022061943,
+      "grad_norm": 0.16240429253875313,
+      "learning_rate": 4.999999560970061e-05,
+      "loss": 1.5488,
+      "step": 590
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "grad_norm": 0.12575424857894482,
+      "learning_rate": 4.999946877563971e-05,
+      "loss": 1.532,
+      "step": 600
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "eval_loss": 1.519254446029663,
+      "eval_runtime": 53.3242,
+      "eval_samples_per_second": 7.145,
+      "eval_steps_per_second": 0.9,
+      "step": 600
+    },
+    {
+      "epoch": 0.5176071277047094,
+      "grad_norm": 0.18688482756329736,
+      "learning_rate": 4.999806390290309e-05,
+      "loss": 1.5544,
+      "step": 610
+    },
+    {
+      "epoch": 0.5260924904539669,
+      "grad_norm": 0.12425469431830571,
+      "learning_rate": 4.999578104083307e-05,
+      "loss": 1.5443,
+      "step": 620
+    },
+    {
+      "epoch": 0.5345778532032245,
+      "grad_norm": 0.1299027485420099,
+      "learning_rate": 4.999262026960902e-05,
+      "loss": 1.5569,
+      "step": 630
+    },
+    {
+      "epoch": 0.543063215952482,
+      "grad_norm": 0.11441754852508934,
+      "learning_rate": 4.998858170024449e-05,
+      "loss": 1.5316,
+      "step": 640
+    },
+    {
+      "epoch": 0.5515485787017395,
+      "grad_norm": 0.14888547248976478,
+      "learning_rate": 4.998366547458326e-05,
+      "loss": 1.5177,
+      "step": 650
+    },
+    {
+      "epoch": 0.560033941450997,
+      "grad_norm": 0.14859292774768867,
+      "learning_rate": 4.997787176529449e-05,
+      "loss": 1.5394,
+      "step": 660
+    },
+    {
+      "epoch": 0.5685193042002545,
+      "grad_norm": 0.12499154376539734,
+      "learning_rate": 4.997120077586651e-05,
+      "loss": 1.5554,
+      "step": 670
+    },
+    {
+      "epoch": 0.5770046669495121,
+      "grad_norm": 0.1218974898058821,
+      "learning_rate": 4.9963652740599774e-05,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.5854900296987696,
+      "grad_norm": 0.1273110498715124,
+      "learning_rate": 4.995522792459859e-05,
+      "loss": 1.5349,
+      "step": 690
+    },
+    {
+      "epoch": 0.5939753924480271,
+      "grad_norm": 0.12115412881719101,
+      "learning_rate": 4.994592662376183e-05,
+      "loss": 1.5419,
+      "step": 700
+    },
+    {
+      "epoch": 0.6024607551972847,
+      "grad_norm": 0.14855096330233286,
+      "learning_rate": 4.99357491647725e-05,
+      "loss": 1.513,
+      "step": 710
+    },
+    {
+      "epoch": 0.6109461179465422,
+      "grad_norm": 0.11407988659327956,
+      "learning_rate": 4.992469590508628e-05,
+      "loss": 1.5243,
+      "step": 720
+    },
+    {
+      "epoch": 0.6194314806957998,
+      "grad_norm": 0.1197712643781127,
+      "learning_rate": 4.9912767232919035e-05,
+      "loss": 1.5177,
+      "step": 730
+    },
+    {
+      "epoch": 0.6279168434450573,
+      "grad_norm": 0.12400515877262065,
+      "learning_rate": 4.9899963567233074e-05,
+      "loss": 1.5619,
+      "step": 740
+    },
+    {
+      "epoch": 0.6364022061943148,
+      "grad_norm": 0.12250385257708406,
+      "learning_rate": 4.988628535772249e-05,
+      "loss": 1.539,
+      "step": 750
+    },
+    {
+      "epoch": 0.6448875689435724,
+      "grad_norm": 0.1262441090496857,
+      "learning_rate": 4.987173308479738e-05,
+      "loss": 1.5195,
+      "step": 760
+    },
+    {
+      "epoch": 0.6533729316928298,
+      "grad_norm": 0.12459694416473029,
+      "learning_rate": 4.985630725956694e-05,
+      "loss": 1.5462,
+      "step": 770
+    },
+    {
+      "epoch": 0.6618582944420874,
+      "grad_norm": 0.12985189006106762,
+      "learning_rate": 4.9840008423821527e-05,
+      "loss": 1.5113,
+      "step": 780
+    },
+    {
+      "epoch": 0.6703436571913449,
+      "grad_norm": 0.12689306141471304,
+      "learning_rate": 4.9822837150013636e-05,
+      "loss": 1.5201,
+      "step": 790
+    },
+    {
+      "epoch": 0.6788290199406024,
+      "grad_norm": 0.15393156370587963,
+      "learning_rate": 4.980479404123778e-05,
+      "loss": 1.5121,
+      "step": 800
+    },
+    {
+      "epoch": 0.68731438268986,
+      "grad_norm": 0.13213701895207608,
+      "learning_rate": 4.978587973120931e-05,
+      "loss": 1.5307,
+      "step": 810
+    },
+    {
+      "epoch": 0.6957997454391175,
+      "grad_norm": 0.11561354931316294,
+      "learning_rate": 4.9766094884242184e-05,
+      "loss": 1.5316,
+      "step": 820
+    },
+    {
+      "epoch": 0.7042851081883751,
+      "grad_norm": 0.12414772399330044,
+      "learning_rate": 4.974544019522559e-05,
+      "loss": 1.5148,
+      "step": 830
+    },
+    {
+      "epoch": 0.7127704709376326,
+      "grad_norm": 0.1171652849153521,
+      "learning_rate": 4.972391638959959e-05,
+      "loss": 1.5096,
+      "step": 840
+    },
+    {
+      "epoch": 0.7212558336868902,
+      "grad_norm": 0.12868937349582316,
+      "learning_rate": 4.9701524223329585e-05,
+      "loss": 1.5282,
+      "step": 850
+    },
+    {
+      "epoch": 0.7297411964361477,
+      "grad_norm": 0.1200015077117309,
+      "learning_rate": 4.967826448287981e-05,
+      "loss": 1.5512,
+      "step": 860
+    },
+    {
+      "epoch": 0.7382265591854051,
+      "grad_norm": 0.12340885660045105,
+      "learning_rate": 4.96541379851857e-05,
+      "loss": 1.5394,
+      "step": 870
+    },
+    {
+      "epoch": 0.7467119219346627,
+      "grad_norm": 0.12976937691467555,
+      "learning_rate": 4.962914557762517e-05,
+      "loss": 1.51,
+      "step": 880
+    },
+    {
+      "epoch": 0.7551972846839202,
+      "grad_norm": 0.11912878476038466,
+      "learning_rate": 4.9603288137988905e-05,
+      "loss": 1.5294,
+      "step": 890
+    },
+    {
+      "epoch": 0.7636826474331778,
+      "grad_norm": 0.1299625480337927,
+      "learning_rate": 4.957656657444947e-05,
+      "loss": 1.507,
+      "step": 900
+    },
+    {
+      "epoch": 0.7721680101824353,
+      "grad_norm": 0.12380144459698468,
+      "learning_rate": 4.954898182552946e-05,
+      "loss": 1.5376,
+      "step": 910
+    },
+    {
+      "epoch": 0.7806533729316928,
+      "grad_norm": 0.13139339643682763,
+      "learning_rate": 4.9520534860068535e-05,
+      "loss": 1.5291,
+      "step": 920
+    },
+    {
+      "epoch": 0.7891387356809504,
+      "grad_norm": 0.13088956203983898,
+      "learning_rate": 4.949122667718935e-05,
+      "loss": 1.5239,
+      "step": 930
+    },
+    {
+      "epoch": 0.7976240984302079,
+      "grad_norm": 0.12586052988453703,
+      "learning_rate": 4.94610583062625e-05,
+      "loss": 1.5525,
+      "step": 940
+    },
+    {
+      "epoch": 0.8061094611794655,
+      "grad_norm": 0.12020996031652877,
+      "learning_rate": 4.943003080687035e-05,
+      "loss": 1.5525,
+      "step": 950
+    },
+    {
+      "epoch": 0.814594823928723,
+      "grad_norm": 0.12866375954060869,
+      "learning_rate": 4.9398145268769856e-05,
+      "loss": 1.5266,
+      "step": 960
+    },
+    {
+      "epoch": 0.8230801866779804,
+      "grad_norm": 0.13166136756817035,
+      "learning_rate": 4.936540281185423e-05,
+      "loss": 1.5041,
+      "step": 970
+    },
+    {
+      "epoch": 0.831565549427238,
+      "grad_norm": 0.12481946698483787,
+      "learning_rate": 4.933180458611364e-05,
+      "loss": 1.5271,
+      "step": 980
+    },
+    {
+      "epoch": 0.8400509121764955,
+      "grad_norm": 0.12264463761209114,
+      "learning_rate": 4.9297351771594844e-05,
+      "loss": 1.5354,
+      "step": 990
+    },
+    {
+      "epoch": 0.8485362749257531,
+      "grad_norm": 0.11985452856537594,
+      "learning_rate": 4.926204557835968e-05,
+      "loss": 1.5167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8570216376750106,
+      "grad_norm": 0.13125396521190327,
+      "learning_rate": 4.9225887246442634e-05,
+      "loss": 1.5282,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8655070004242681,
+      "grad_norm": 0.12730192328072554,
+      "learning_rate": 4.918887804580725e-05,
+      "loss": 1.5089,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8739923631735257,
+      "grad_norm": 0.12724644219344786,
+      "learning_rate": 4.915101927630153e-05,
+      "loss": 1.4964,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8824777259227832,
+      "grad_norm": 0.13578611501833232,
+      "learning_rate": 4.911231226761227e-05,
+      "loss": 1.5189,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8909630886720408,
+      "grad_norm": 0.13577513964986457,
+      "learning_rate": 4.90727583792184e-05,
+      "loss": 1.5149,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8994484514212983,
+      "grad_norm": 0.1269735011676505,
+      "learning_rate": 4.903235900034317e-05,
+      "loss": 1.5066,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9079338141705557,
+      "grad_norm": 0.13250058214235566,
+      "learning_rate": 4.899111554990543e-05,
+      "loss": 1.5129,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9164191769198133,
+      "grad_norm": 0.13130735246433495,
+      "learning_rate": 4.894902947646975e-05,
+      "loss": 1.5156,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9249045396690708,
+      "grad_norm": 0.1273580180253049,
+      "learning_rate": 4.890610225819553e-05,
+      "loss": 1.5324,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9333899024183284,
+      "grad_norm": 0.13155314243939242,
+      "learning_rate": 4.8862335402785136e-05,
+      "loss": 1.5106,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9418752651675859,
+      "grad_norm": 0.13564895211984299,
+      "learning_rate": 4.88177304474309e-05,
+      "loss": 1.5067,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9503606279168434,
+      "grad_norm": 0.12774735587114736,
+      "learning_rate": 4.877228895876115e-05,
+      "loss": 1.5182,
+      "step": 1120
+    },
+    {
+      "epoch": 0.958845990666101,
+      "grad_norm": 0.1307997709537685,
+      "learning_rate": 4.872601253278517e-05,
+      "loss": 1.4969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9673313534153585,
+      "grad_norm": 0.1304794845040634,
+      "learning_rate": 4.867890279483717e-05,
+      "loss": 1.5264,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9758167161646161,
+      "grad_norm": 0.13666141796489684,
+      "learning_rate": 4.8630961399519206e-05,
+      "loss": 1.5467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9843020789138736,
+      "grad_norm": 0.1370278303190263,
+      "learning_rate": 4.8582190030643e-05,
+      "loss": 1.5127,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9927874416631312,
+      "grad_norm": 0.1390936629299565,
+      "learning_rate": 4.8532590401170894e-05,
+      "loss": 1.5058,
+      "step": 1170
+    },
+    {
+      "epoch": 1.0012728044123886,
+      "grad_norm": 0.12934475548108287,
+      "learning_rate": 4.848216425315561e-05,
+      "loss": 1.5202,
+      "step": 1180
+    },
+    {
+      "epoch": 1.0097581671616462,
+      "grad_norm": 0.13898591683370803,
+      "learning_rate": 4.843091335767913e-05,
+      "loss": 1.4563,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "grad_norm": 0.17488231535826249,
+      "learning_rate": 4.837883951479043e-05,
+      "loss": 1.4402,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "eval_loss": 1.4955657720565796,
+      "eval_runtime": 52.424,
+      "eval_samples_per_second": 7.268,
+      "eval_steps_per_second": 0.916,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0267288926601612,
+      "grad_norm": 0.1536036344095855,
+      "learning_rate": 4.832594455344229e-05,
+      "loss": 1.4848,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0352142554094188,
+      "grad_norm": 0.15762414421336599,
+      "learning_rate": 4.827223033142706e-05,
+      "loss": 1.4567,
+      "step": 1220
+    },
+    {
+      "epoch": 1.0436996181586762,
+      "grad_norm": 0.15058229398130366,
+      "learning_rate": 4.8217698735311414e-05,
+      "loss": 1.4672,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0521849809079338,
+      "grad_norm": 0.16010992835678386,
+      "learning_rate": 4.8162351680370044e-05,
+      "loss": 1.4458,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0606703436571914,
+      "grad_norm": 0.16758816000341356,
+      "learning_rate": 4.810619111051847e-05,
+      "loss": 1.4842,
+      "step": 1250
+    },
+    {
+      "epoch": 1.069155706406449,
+      "grad_norm": 0.16559260972674986,
+      "learning_rate": 4.8049218998244696e-05,
+      "loss": 1.4556,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0776410691557063,
+      "grad_norm": 0.17237632034416966,
+      "learning_rate": 4.7991437344539966e-05,
+      "loss": 1.4813,
+      "step": 1270
+    },
+    {
+      "epoch": 1.086126431904964,
+      "grad_norm": 0.17112756741722487,
+      "learning_rate": 4.793284817882845e-05,
+      "loss": 1.4535,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0946117946542215,
+      "grad_norm": 0.16828572707718548,
+      "learning_rate": 4.787345355889604e-05,
+      "loss": 1.4344,
+      "step": 1290
+    },
+    {
+      "epoch": 1.103097157403479,
+      "grad_norm": 0.15709986047041227,
+      "learning_rate": 4.7813255570817985e-05,
+      "loss": 1.4744,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1115825201527365,
+      "grad_norm": 0.16651547128146313,
+      "learning_rate": 4.775225632888568e-05,
+      "loss": 1.4561,
+      "step": 1310
+    },
+    {
+      "epoch": 1.120067882901994,
+      "grad_norm": 0.16750176017515714,
+      "learning_rate": 4.76904579755324e-05,
+      "loss": 1.4616,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1285532456512515,
+      "grad_norm": 0.1608016567554825,
+      "learning_rate": 4.7627862681258037e-05,
+      "loss": 1.4593,
+      "step": 1330
+    },
+    {
+      "epoch": 1.137038608400509,
+      "grad_norm": 0.21390766919038295,
+      "learning_rate": 4.756447264455287e-05,
+      "loss": 1.4484,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1455239711497667,
+      "grad_norm": 0.16826883293172662,
+      "learning_rate": 4.750029009182038e-05,
+      "loss": 1.4703,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1540093338990243,
+      "grad_norm": 0.17431508867079595,
+      "learning_rate": 4.7435317277299e-05,
+      "loss": 1.4701,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1624946966482816,
+      "grad_norm": 0.15973851467570443,
+      "learning_rate": 4.736955648298299e-05,
+      "loss": 1.4503,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1709800593975392,
+      "grad_norm": 0.1887713767970947,
+      "learning_rate": 4.730301001854225e-05,
+      "loss": 1.4624,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1794654221467968,
+      "grad_norm": 0.16898695344997974,
+      "learning_rate": 4.7235680221241216e-05,
+      "loss": 1.4452,
+      "step": 1390
+    },
+    {
+      "epoch": 1.1879507848960542,
+      "grad_norm": 0.20014553287073528,
+      "learning_rate": 4.716756945585681e-05,
+      "loss": 1.4717,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1964361476453118,
+      "grad_norm": 0.17137954325200072,
+      "learning_rate": 4.709868011459528e-05,
+      "loss": 1.4403,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2049215103945694,
+      "grad_norm": 0.17801721751888322,
+      "learning_rate": 4.7029014617008294e-05,
+      "loss": 1.4339,
+      "step": 1420
+    },
+    {
+      "epoch": 1.213406873143827,
+      "grad_norm": 0.17139613676642362,
+      "learning_rate": 4.695857540990789e-05,
+      "loss": 1.4573,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2218922358930844,
+      "grad_norm": 0.16971403514498054,
+      "learning_rate": 4.688736496728058e-05,
+      "loss": 1.4282,
+      "step": 1440
+    },
+    {
+      "epoch": 1.230377598642342,
+      "grad_norm": 0.17200272420880428,
+      "learning_rate": 4.681538579020038e-05,
+      "loss": 1.4434,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2388629613915996,
+      "grad_norm": 0.17208160407432616,
+      "learning_rate": 4.6742640406741106e-05,
+      "loss": 1.45,
+      "step": 1460
+    },
+    {
+      "epoch": 1.247348324140857,
+      "grad_norm": 0.1939626212901777,
+      "learning_rate": 4.666913137188743e-05,
+      "loss": 1.4608,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2558336868901145,
+      "grad_norm": 0.17291794493304186,
+      "learning_rate": 4.6594861267445236e-05,
+      "loss": 1.4671,
+      "step": 1480
+    },
+    {
+      "epoch": 1.2643190496393721,
+      "grad_norm": 0.18219792041638924,
+      "learning_rate": 4.651983270195093e-05,
+      "loss": 1.4262,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2728044123886297,
+      "grad_norm": 0.18086437830489926,
+      "learning_rate": 4.644404831057979e-05,
+      "loss": 1.4455,
+      "step": 1500
+    },
+    {
+      "epoch": 1.281289775137887,
+      "grad_norm": 0.17417619624549402,
+      "learning_rate": 4.636751075505344e-05,
+      "loss": 1.4873,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2897751378871447,
+      "grad_norm": 0.18354282411845188,
+      "learning_rate": 4.629022272354637e-05,
+      "loss": 1.4525,
+      "step": 1520
+    },
+    {
+      "epoch": 1.298260500636402,
+      "grad_norm": 0.17985617345325455,
+      "learning_rate": 4.621218693059149e-05,
+      "loss": 1.4303,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3067458633856597,
+      "grad_norm": 0.1809708317849863,
+      "learning_rate": 4.6133406116984795e-05,
+      "loss": 1.4631,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3152312261349173,
+      "grad_norm": 0.17487374671212322,
+      "learning_rate": 4.6053883049689145e-05,
+      "loss": 1.4482,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3237165888841749,
+      "grad_norm": 0.19912807671077193,
+      "learning_rate": 4.5973620521737036e-05,
+      "loss": 1.4497,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3322019516334322,
+      "grad_norm": 0.17853627546912074,
+      "learning_rate": 4.5892621352132514e-05,
+      "loss": 1.4456,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3406873143826898,
+      "grad_norm": 0.18252596927754394,
+      "learning_rate": 4.581088838575218e-05,
+      "loss": 1.4328,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3491726771319474,
+      "grad_norm": 0.17604951053556211,
+      "learning_rate": 4.572842449324525e-05,
+      "loss": 1.4442,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3576580398812048,
+      "grad_norm": 0.18358942463311748,
+      "learning_rate": 4.564523257093275e-05,
+      "loss": 1.4338,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3661434026304624,
+      "grad_norm": 0.20508703236267142,
+      "learning_rate": 4.5561315540705774e-05,
+      "loss": 1.4445,
+      "step": 1610
+    },
+    {
+      "epoch": 1.37462876537972,
+      "grad_norm": 0.18486352550747187,
+      "learning_rate": 4.547667634992288e-05,
+      "loss": 1.4261,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3831141281289776,
+      "grad_norm": 0.17492766465456316,
+      "learning_rate": 4.539131797130656e-05,
+      "loss": 1.4258,
+      "step": 1630
+    },
+    {
+      "epoch": 1.391599490878235,
+      "grad_norm": 0.19692876587833674,
+      "learning_rate": 4.530524340283881e-05,
+      "loss": 1.4349,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4000848536274926,
+      "grad_norm": 0.19155373430892478,
+      "learning_rate": 4.521845566765589e-05,
+      "loss": 1.4536,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4085702163767502,
+      "grad_norm": 0.18544325977459192,
+      "learning_rate": 4.513095781394208e-05,
+      "loss": 1.4363,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4170555791260075,
+      "grad_norm": 0.177828004720666,
+      "learning_rate": 4.504275291482267e-05,
+      "loss": 1.4595,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4255409418752651,
+      "grad_norm": 0.17855432230356816,
+      "learning_rate": 4.495384406825601e-05,
+      "loss": 1.4211,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4340263046245227,
+      "grad_norm": 0.20232492538380317,
+      "learning_rate": 4.486423439692469e-05,
+      "loss": 1.4189,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4425116673737803,
+      "grad_norm": 0.1975109303350431,
+      "learning_rate": 4.477392704812585e-05,
+      "loss": 1.4565,
+      "step": 1700
+    },
+    {
+      "epoch": 1.4509970301230377,
+      "grad_norm": 0.19619010830399825,
+      "learning_rate": 4.468292519366071e-05,
+      "loss": 1.4382,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4594823928722953,
+      "grad_norm": 0.18168826428246143,
+      "learning_rate": 4.459123202972308e-05,
+      "loss": 1.4471,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4679677556215527,
+      "grad_norm": 0.1923264062362399,
+      "learning_rate": 4.449885077678717e-05,
+      "loss": 1.4153,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4764531183708103,
+      "grad_norm": 0.1907937313040222,
+      "learning_rate": 4.440578467949445e-05,
+      "loss": 1.4432,
+      "step": 1740
+    },
+    {
+      "epoch": 1.4849384811200679,
+      "grad_norm": 0.19107457667767244,
+      "learning_rate": 4.431203700653968e-05,
+      "loss": 1.4285,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4934238438693255,
+      "grad_norm": 0.19847350429107552,
+      "learning_rate": 4.421761105055613e-05,
+      "loss": 1.4383,
+      "step": 1760
+    },
+    {
+      "epoch": 1.501909206618583,
+      "grad_norm": 0.18536475556610216,
+      "learning_rate": 4.4122510127999937e-05,
+      "loss": 1.42,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5103945693678404,
+      "grad_norm": 0.18481023473586697,
+      "learning_rate": 4.4026737579033584e-05,
+      "loss": 1.4384,
+      "step": 1780
+    },
+    {
+      "epoch": 1.518879932117098,
+      "grad_norm": 0.20863867505874642,
+      "learning_rate": 4.393029676740864e-05,
+      "loss": 1.4543,
+      "step": 1790
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "grad_norm": 0.1816036870853105,
+      "learning_rate": 4.3833191080347575e-05,
+      "loss": 1.434,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "eval_loss": 1.4622184038162231,
+      "eval_runtime": 52.4041,
+      "eval_samples_per_second": 7.27,
+      "eval_steps_per_second": 0.916,
+      "step": 1800
+    },
+    {
+      "epoch": 1.535850657615613,
+      "grad_norm": 0.19378252368958881,
+      "learning_rate": 4.3735423928424815e-05,
+      "loss": 1.4275,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5443360203648706,
+      "grad_norm": 0.20453331251433848,
+      "learning_rate": 4.363699874544697e-05,
+      "loss": 1.4203,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5528213831141282,
+      "grad_norm": 0.26684319417219377,
+      "learning_rate": 4.3537918988332156e-05,
+      "loss": 1.4372,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5613067458633858,
+      "grad_norm": 0.25745160303419773,
+      "learning_rate": 4.343818813698868e-05,
+      "loss": 1.4082,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5697921086126432,
+      "grad_norm": 0.19969727996700776,
+      "learning_rate": 4.3337809694192765e-05,
+      "loss": 1.4314,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5782774713619008,
+      "grad_norm": 0.20117210832277968,
+      "learning_rate": 4.3236787185465525e-05,
+      "loss": 1.4293,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5867628341111581,
+      "grad_norm": 0.20173003641028897,
+      "learning_rate": 4.313512415894913e-05,
+      "loss": 1.4406,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5952481968604157,
+      "grad_norm": 0.20304770794371527,
+      "learning_rate": 4.303282418528224e-05,
+      "loss": 1.4286,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6037335596096733,
+      "grad_norm": 0.19126658907738198,
+      "learning_rate": 4.292989085747452e-05,
+      "loss": 1.4184,
+      "step": 1890
+    },
+    {
+      "epoch": 1.612218922358931,
+      "grad_norm": 0.20069554966453027,
+      "learning_rate": 4.282632779078051e-05,
+      "loss": 1.4133,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6207042851081885,
+      "grad_norm": 0.1952881519566686,
+      "learning_rate": 4.2722138622572624e-05,
+      "loss": 1.4432,
+      "step": 1910
+    },
+    {
+      "epoch": 1.629189647857446,
+      "grad_norm": 0.19763704668680288,
+      "learning_rate": 4.261732701221339e-05,
+      "loss": 1.3921,
+      "step": 1920
+    },
+    {
+      "epoch": 1.6376750106067033,
+      "grad_norm": 0.19821464294464497,
+      "learning_rate": 4.2511896640926925e-05,
+      "loss": 1.4454,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6461603733559609,
+      "grad_norm": 0.20456545626297834,
+      "learning_rate": 4.240585121166966e-05,
+      "loss": 1.4147,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6546457361052185,
+      "grad_norm": 0.2119092529186395,
+      "learning_rate": 4.229919444900027e-05,
+      "loss": 1.3969,
+      "step": 1950
+    },
+    {
+      "epoch": 1.663131098854476,
+      "grad_norm": 0.20330157582122357,
+      "learning_rate": 4.2191930098948865e-05,
+      "loss": 1.426,
+      "step": 1960
+    },
+    {
+      "epoch": 1.6716164616037337,
+      "grad_norm": 0.21761164739298738,
+      "learning_rate": 4.2084061928885406e-05,
+      "loss": 1.4246,
+      "step": 1970
+    },
+    {
+      "epoch": 1.680101824352991,
+      "grad_norm": 0.19331588142071401,
+      "learning_rate": 4.197559372738741e-05,
+      "loss": 1.4305,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6885871871022486,
+      "grad_norm": 0.20188460724329996,
+      "learning_rate": 4.186652930410685e-05,
+      "loss": 1.4153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.697072549851506,
+      "grad_norm": 0.20988950033571588,
+      "learning_rate": 4.1756872489636425e-05,
+      "loss": 1.3894,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7055579126007636,
+      "grad_norm": 0.1966475893123187,
+      "learning_rate": 4.1646627135374916e-05,
+      "loss": 1.3962,
+      "step": 2010
+    },
+    {
+      "epoch": 1.7140432753500212,
+      "grad_norm": 0.20785207367991768,
+      "learning_rate": 4.1535797113392004e-05,
+      "loss": 1.4037,
+      "step": 2020
+    },
+    {
+      "epoch": 1.7225286380992788,
+      "grad_norm": 0.2029940281663133,
+      "learning_rate": 4.1424386316292224e-05,
+      "loss": 1.4011,
+      "step": 2030
+    },
+    {
+      "epoch": 1.7310140008485364,
+      "grad_norm": 0.2247844551379277,
+      "learning_rate": 4.131239865707829e-05,
+      "loss": 1.4084,
+      "step": 2040
+    },
+    {
+      "epoch": 1.7394993635977938,
+      "grad_norm": 0.20900441746105022,
+      "learning_rate": 4.11998380690136e-05,
+      "loss": 1.4235,
+      "step": 2050
+    },
+    {
+      "epoch": 1.7479847263470514,
+      "grad_norm": 0.20362408546889926,
+      "learning_rate": 4.108670850548416e-05,
+      "loss": 1.4204,
+      "step": 2060
+    },
+    {
+      "epoch": 1.7564700890963088,
+      "grad_norm": 0.22281567946240438,
+      "learning_rate": 4.097301393985968e-05,
+      "loss": 1.4023,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7649554518455663,
+      "grad_norm": 0.20867113178797225,
+      "learning_rate": 4.085875836535404e-05,
+      "loss": 1.3895,
+      "step": 2080
+    },
+    {
+      "epoch": 1.773440814594824,
+      "grad_norm": 0.22113231886160947,
+      "learning_rate": 4.0743945794885063e-05,
+      "loss": 1.3963,
+      "step": 2090
+    },
+    {
+      "epoch": 1.7819261773440815,
+      "grad_norm": 0.22334563577844263,
+      "learning_rate": 4.062858026093351e-05,
+      "loss": 1.3988,
+      "step": 2100
+    },
+    {
+      "epoch": 1.7904115400933391,
+      "grad_norm": 0.23218581668265403,
+      "learning_rate": 4.051266581540152e-05,
+      "loss": 1.4068,
+      "step": 2110
+    },
+    {
+      "epoch": 1.7988969028425965,
+      "grad_norm": 0.20295589384571033,
+      "learning_rate": 4.0396206529470234e-05,
+      "loss": 1.3883,
+      "step": 2120
+    },
+    {
+      "epoch": 1.8073822655918539,
+      "grad_norm": 0.22861611442392848,
+      "learning_rate": 4.027920649345687e-05,
+      "loss": 1.4043,
+      "step": 2130
+    },
+    {
+      "epoch": 1.8158676283411115,
+      "grad_norm": 0.2083012771089638,
+      "learning_rate": 4.0161669816671e-05,
+      "loss": 1.398,
+      "step": 2140
+    },
+    {
+      "epoch": 1.824352991090369,
+      "grad_norm": 0.21936173231840464,
+      "learning_rate": 4.004360062727028e-05,
+      "loss": 1.4142,
+      "step": 2150
+    },
+    {
+      "epoch": 1.8328383538396267,
+      "grad_norm": 0.21383435796328337,
+      "learning_rate": 3.9925003072115406e-05,
+      "loss": 1.4138,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8413237165888843,
+      "grad_norm": 0.23301608248270392,
+      "learning_rate": 3.9805881316624506e-05,
+      "loss": 1.4195,
+      "step": 2170
+    },
+    {
+      "epoch": 1.8498090793381419,
+      "grad_norm": 0.22424766656883474,
+      "learning_rate": 3.968623954462681e-05,
+      "loss": 1.4011,
+      "step": 2180
+    },
+    {
+      "epoch": 1.8582944420873992,
+      "grad_norm": 0.21286417342881453,
+      "learning_rate": 3.9566081958215734e-05,
+      "loss": 1.409,
+      "step": 2190
+    },
+    {
+      "epoch": 1.8667798048366566,
+      "grad_norm": 0.21944800687444807,
+      "learning_rate": 3.9445412777601284e-05,
+      "loss": 1.3877,
+      "step": 2200
+    },
+    {
+      "epoch": 1.8752651675859142,
+      "grad_norm": 0.23113173625974803,
+      "learning_rate": 3.932423624096181e-05,
+      "loss": 1.4089,
+      "step": 2210
+    },
+    {
+      "epoch": 1.8837505303351718,
+      "grad_norm": 0.2081941699587778,
+      "learning_rate": 3.920255660429517e-05,
+      "loss": 1.4024,
+      "step": 2220
+    },
+    {
+      "epoch": 1.8922358930844294,
+      "grad_norm": 0.2188685806654701,
+      "learning_rate": 3.908037814126927e-05,
+      "loss": 1.3878,
+      "step": 2230
+    },
+    {
+      "epoch": 1.900721255833687,
+      "grad_norm": 0.22761843244757962,
+      "learning_rate": 3.895770514307193e-05,
+      "loss": 1.4004,
+      "step": 2240
+    },
+    {
+      "epoch": 1.9092066185829444,
+      "grad_norm": 0.23309183623120422,
+      "learning_rate": 3.883454191826017e-05,
+      "loss": 1.4188,
+      "step": 2250
+    },
+    {
+      "epoch": 1.917691981332202,
+      "grad_norm": 0.20329785843911802,
+      "learning_rate": 3.871089279260891e-05,
+      "loss": 1.3893,
+      "step": 2260
+    },
+    {
+      "epoch": 1.9261773440814594,
+      "grad_norm": 0.23470973193726366,
+      "learning_rate": 3.8586762108958995e-05,
+      "loss": 1.3974,
+      "step": 2270
+    },
+    {
+      "epoch": 1.934662706830717,
+      "grad_norm": 0.22779136837044714,
+      "learning_rate": 3.8462154227064725e-05,
+      "loss": 1.4115,
+      "step": 2280
+    },
+    {
+      "epoch": 1.9431480695799745,
+      "grad_norm": 0.22338952315651892,
+      "learning_rate": 3.833707352344068e-05,
+      "loss": 1.3873,
+      "step": 2290
+    },
+    {
+      "epoch": 1.9516334323292321,
+      "grad_norm": 0.23069304025882129,
+      "learning_rate": 3.821152439120801e-05,
+      "loss": 1.3944,
+      "step": 2300
+    },
+    {
+      "epoch": 1.9601187950784897,
+      "grad_norm": 0.23590596270163203,
+      "learning_rate": 3.808551123994018e-05,
+      "loss": 1.3857,
+      "step": 2310
+    },
+    {
+      "epoch": 1.9686041578277471,
+      "grad_norm": 0.22545661808214923,
+      "learning_rate": 3.795903849550805e-05,
+      "loss": 1.3628,
+      "step": 2320
+    },
+    {
+      "epoch": 1.9770895205770047,
+      "grad_norm": 0.2450769875954842,
+      "learning_rate": 3.7832110599924455e-05,
+      "loss": 1.4079,
+      "step": 2330
+    },
+    {
+      "epoch": 1.985574883326262,
+      "grad_norm": 0.22931499326784313,
+      "learning_rate": 3.7704732011188166e-05,
+      "loss": 1.379,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9940602460755197,
+      "grad_norm": 0.22417244507397657,
+      "learning_rate": 3.7576907203127346e-05,
+      "loss": 1.4035,
+      "step": 2350
+    },
+    {
+      "epoch": 2.0025456088247773,
+      "grad_norm": 0.24496197221575314,
+      "learning_rate": 3.7448640665242406e-05,
+      "loss": 1.442,
+      "step": 2360
+    },
+    {
+      "epoch": 2.011030971574035,
+      "grad_norm": 0.2532740296990078,
+      "learning_rate": 3.73199369025483e-05,
+      "loss": 1.2672,
+      "step": 2370
+    },
+    {
+      "epoch": 2.0195163343232925,
+      "grad_norm": 0.2890155987968593,
+      "learning_rate": 3.7190800435416355e-05,
+      "loss": 1.246,
+      "step": 2380
+    },
+    {
+      "epoch": 2.02800169707255,
+      "grad_norm": 0.2541972565696406,
+      "learning_rate": 3.706123579941545e-05,
+      "loss": 1.2603,
+      "step": 2390
+    },
+    {
+      "epoch": 2.036487059821807,
+      "grad_norm": 0.2530140862527023,
+      "learning_rate": 3.693124754515272e-05,
+      "loss": 1.2638,
+      "step": 2400
+    },
+    {
+      "epoch": 2.036487059821807,
+      "eval_loss": 1.435962438583374,
+      "eval_runtime": 52.582,
+      "eval_samples_per_second": 7.246,
+      "eval_steps_per_second": 0.913,
+      "step": 2400
+    },
+    {
+      "epoch": 2.044972422571065,
+      "grad_norm": 0.25100458343337734,
+      "learning_rate": 3.680084023811377e-05,
+      "loss": 1.2711,
+      "step": 2410
+    },
+    {
+      "epoch": 2.0534577853203224,
+      "grad_norm": 0.2695727673292618,
+      "learning_rate": 3.66700184585023e-05,
+      "loss": 1.2578,
+      "step": 2420
+    },
+    {
+      "epoch": 2.06194314806958,
+      "grad_norm": 0.2605068415443213,
+      "learning_rate": 3.6538786801079226e-05,
+      "loss": 1.2506,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0704285108188376,
+      "grad_norm": 0.27415607207865045,
+      "learning_rate": 3.64071498750013e-05,
+      "loss": 1.2852,
+      "step": 2440
+    },
+    {
+      "epoch": 2.078913873568095,
+      "grad_norm": 0.2688900338206285,
+      "learning_rate": 3.627511230365928e-05,
+      "loss": 1.2695,
+      "step": 2450
+    },
+    {
+      "epoch": 2.0873992363173524,
+      "grad_norm": 0.2750825805336503,
+      "learning_rate": 3.614267872451546e-05,
+      "loss": 1.2643,
+      "step": 2460
+    },
+    {
+      "epoch": 2.09588459906661,
+      "grad_norm": 0.2659269066581903,
+      "learning_rate": 3.600985378894086e-05,
+      "loss": 1.2868,
+      "step": 2470
+    },
+    {
+      "epoch": 2.1043699618158676,
+      "grad_norm": 0.24411151291321526,
+      "learning_rate": 3.587664216205183e-05,
+      "loss": 1.2571,
+      "step": 2480
+    },
+    {
+      "epoch": 2.112855324565125,
+      "grad_norm": 0.2574194755634052,
+      "learning_rate": 3.574304852254621e-05,
+      "loss": 1.2769,
+      "step": 2490
+    },
+    {
+      "epoch": 2.1213406873143827,
+      "grad_norm": 0.2894545074998905,
+      "learning_rate": 3.5609077562538997e-05,
+      "loss": 1.2469,
+      "step": 2500
+    },
+    {
+      "epoch": 2.1298260500636403,
+      "grad_norm": 0.2828176429904294,
+      "learning_rate": 3.547473398739754e-05,
+      "loss": 1.2527,
+      "step": 2510
+    },
+    {
+      "epoch": 2.138311412812898,
+      "grad_norm": 0.25886029771650565,
+      "learning_rate": 3.5340022515576294e-05,
+      "loss": 1.2578,
+      "step": 2520
+    },
+    {
+      "epoch": 2.146796775562155,
+      "grad_norm": 0.2783799371621383,
+      "learning_rate": 3.52049478784511e-05,
+      "loss": 1.2489,
+      "step": 2530
+    },
+    {
+      "epoch": 2.1552821383114127,
+      "grad_norm": 0.2753116113218978,
+      "learning_rate": 3.506951482015297e-05,
+      "loss": 1.275,
+      "step": 2540
+    },
+    {
+      "epoch": 2.1637675010606703,
+      "grad_norm": 0.28115792079727675,
+      "learning_rate": 3.493372809740152e-05,
+      "loss": 1.2554,
+      "step": 2550
+    },
+    {
+      "epoch": 2.172252863809928,
+      "grad_norm": 0.27954425325951715,
+      "learning_rate": 3.479759247933785e-05,
+      "loss": 1.2618,
+      "step": 2560
+    },
+    {
+      "epoch": 2.1807382265591855,
+      "grad_norm": 0.27555174232347995,
+      "learning_rate": 3.466111274735707e-05,
+      "loss": 1.2598,
+      "step": 2570
+    },
+    {
+      "epoch": 2.189223589308443,
+      "grad_norm": 0.27280827991301104,
+      "learning_rate": 3.452429369494037e-05,
+      "loss": 1.262,
+      "step": 2580
+    },
+    {
+      "epoch": 2.1977089520577007,
+      "grad_norm": 0.2749685805551003,
+      "learning_rate": 3.438714012748664e-05,
+      "loss": 1.2683,
+      "step": 2590
+    },
+    {
+      "epoch": 2.206194314806958,
+      "grad_norm": 0.2780594302788235,
+      "learning_rate": 3.424965686214371e-05,
+      "loss": 1.2462,
+      "step": 2600
+    },
+    {
+      "epoch": 2.2146796775562154,
+      "grad_norm": 0.2942257416636676,
+      "learning_rate": 3.411184872763915e-05,
+      "loss": 1.2581,
+      "step": 2610
+    },
+    {
+      "epoch": 2.223165040305473,
+      "grad_norm": 0.27000377333423803,
+      "learning_rate": 3.39737205641107e-05,
+      "loss": 1.2412,
+      "step": 2620
+    },
+    {
+      "epoch": 2.2316504030547306,
+      "grad_norm": 0.28187507810449336,
+      "learning_rate": 3.383527722293622e-05,
+      "loss": 1.2659,
+      "step": 2630
+    },
+    {
+      "epoch": 2.240135765803988,
+      "grad_norm": 0.2736213940552268,
+      "learning_rate": 3.369652356656336e-05,
+      "loss": 1.2553,
+      "step": 2640
+    },
+    {
+      "epoch": 2.248621128553246,
+      "grad_norm": 0.29698834543438446,
+      "learning_rate": 3.355746446833873e-05,
+      "loss": 1.2714,
+      "step": 2650
+    },
+    {
+      "epoch": 2.257106491302503,
+      "grad_norm": 0.2875128112484735,
+      "learning_rate": 3.3418104812336786e-05,
+      "loss": 1.2508,
+      "step": 2660
+    },
+    {
+      "epoch": 2.2655918540517606,
+      "grad_norm": 0.3016647299373059,
+      "learning_rate": 3.327844949318824e-05,
+      "loss": 1.2451,
+      "step": 2670
+    },
+    {
+      "epoch": 2.274077216801018,
+      "grad_norm": 0.27371321581702696,
+      "learning_rate": 3.3138503415908176e-05,
+      "loss": 1.2467,
+      "step": 2680
+    },
+    {
+      "epoch": 2.2825625795502757,
+      "grad_norm": 0.28374547760120017,
+      "learning_rate": 3.299827149572376e-05,
+      "loss": 1.2452,
+      "step": 2690
+    },
+    {
+      "epoch": 2.2910479422995333,
+      "grad_norm": 0.2805999278165284,
+      "learning_rate": 3.285775865790166e-05,
+      "loss": 1.2595,
+      "step": 2700
+    },
+    {
+      "epoch": 2.299533305048791,
+      "grad_norm": 0.2758019804125597,
+      "learning_rate": 3.271696983757496e-05,
+      "loss": 1.2583,
+      "step": 2710
+    },
+    {
+      "epoch": 2.3080186677980485,
+      "grad_norm": 0.27211127699988974,
+      "learning_rate": 3.2575909979569906e-05,
+      "loss": 1.2255,
+      "step": 2720
+    },
+    {
+      "epoch": 2.316504030547306,
+      "grad_norm": 0.2741831859110416,
+      "learning_rate": 3.243458403823223e-05,
+      "loss": 1.2335,
+      "step": 2730
+    },
+    {
+      "epoch": 2.3249893932965633,
+      "grad_norm": 0.287074507507,
+      "learning_rate": 3.2292996977253075e-05,
+      "loss": 1.2555,
+      "step": 2740
+    },
+    {
+      "epoch": 2.333474756045821,
+      "grad_norm": 0.2760197579958247,
+      "learning_rate": 3.215115376949474e-05,
+      "loss": 1.2574,
+      "step": 2750
+    },
+    {
+      "epoch": 2.3419601187950785,
+      "grad_norm": 0.29917391348714156,
+      "learning_rate": 3.200905939681599e-05,
+      "loss": 1.2232,
+      "step": 2760
+    },
+    {
+      "epoch": 2.350445481544336,
+      "grad_norm": 0.2863180346672473,
+      "learning_rate": 3.1866718849897044e-05,
+      "loss": 1.2341,
+      "step": 2770
+    },
+    {
+      "epoch": 2.3589308442935937,
+      "grad_norm": 0.2760526831444543,
+      "learning_rate": 3.172413712806435e-05,
+      "loss": 1.253,
+      "step": 2780
+    },
+    {
+      "epoch": 2.3674162070428513,
+      "grad_norm": 0.29286413736773825,
+      "learning_rate": 3.158131923911498e-05,
+      "loss": 1.2617,
+      "step": 2790
+    },
+    {
+      "epoch": 2.3759015697921084,
+      "grad_norm": 0.27643034174892955,
+      "learning_rate": 3.143827019914072e-05,
+      "loss": 1.2152,
+      "step": 2800
+    },
+    {
+      "epoch": 2.384386932541366,
+      "grad_norm": 0.2939949433037669,
+      "learning_rate": 3.12949950323519e-05,
+      "loss": 1.2354,
+      "step": 2810
+    },
+    {
+      "epoch": 2.3928722952906236,
+      "grad_norm": 0.2864245267570891,
+      "learning_rate": 3.115149877090097e-05,
+      "loss": 1.2447,
+      "step": 2820
+    },
+    {
+      "epoch": 2.401357658039881,
+      "grad_norm": 0.2952829920235313,
+      "learning_rate": 3.1007786454705724e-05,
+      "loss": 1.2462,
+      "step": 2830
+    },
+    {
+      "epoch": 2.409843020789139,
+      "grad_norm": 0.3032080033620836,
+      "learning_rate": 3.0863863131272265e-05,
+      "loss": 1.2317,
+      "step": 2840
+    },
+    {
+      "epoch": 2.4183283835383964,
+      "grad_norm": 0.2678380639415362,
+      "learning_rate": 3.07197338555178e-05,
+      "loss": 1.2466,
+      "step": 2850
+    },
+    {
+      "epoch": 2.426813746287654,
+      "grad_norm": 0.3000338098809928,
+      "learning_rate": 3.0575403689593016e-05,
+      "loss": 1.2469,
+      "step": 2860
+    },
+    {
+      "epoch": 2.435299109036911,
+      "grad_norm": 0.2885428511714088,
+      "learning_rate": 3.043087770270435e-05,
+      "loss": 1.241,
+      "step": 2870
+    },
+    {
+      "epoch": 2.4437844717861688,
+      "grad_norm": 0.2902606566366597,
+      "learning_rate": 3.0286160970935906e-05,
+      "loss": 1.2498,
+      "step": 2880
+    },
+    {
+      "epoch": 2.4522698345354264,
+      "grad_norm": 0.2930924599960876,
+      "learning_rate": 3.0141258577071184e-05,
+      "loss": 1.2508,
+      "step": 2890
+    },
+    {
+      "epoch": 2.460755197284684,
+      "grad_norm": 0.28761403953538467,
+      "learning_rate": 2.9996175610414572e-05,
+      "loss": 1.2379,
+      "step": 2900
+    },
+    {
+      "epoch": 2.4692405600339415,
+      "grad_norm": 0.28888693356528744,
+      "learning_rate": 2.9850917166612586e-05,
+      "loss": 1.2383,
+      "step": 2910
+    },
+    {
+      "epoch": 2.477725922783199,
+      "grad_norm": 0.29714323219094924,
+      "learning_rate": 2.9705488347474896e-05,
+      "loss": 1.2221,
+      "step": 2920
+    },
+    {
+      "epoch": 2.4862112855324563,
+      "grad_norm": 0.3024332099011336,
+      "learning_rate": 2.9559894260795144e-05,
+      "loss": 1.2417,
+      "step": 2930
+    },
+    {
+      "epoch": 2.494696648281714,
+      "grad_norm": 0.2900123354730048,
+      "learning_rate": 2.9414140020171554e-05,
+      "loss": 1.2543,
+      "step": 2940
+    },
+    {
+      "epoch": 2.5031820110309715,
+      "grad_norm": 0.30122390943433014,
+      "learning_rate": 2.926823074482733e-05,
+      "loss": 1.2542,
+      "step": 2950
+    },
+    {
+      "epoch": 2.511667373780229,
+      "grad_norm": 0.2860208265471049,
+      "learning_rate": 2.912217155943083e-05,
+      "loss": 1.2335,
+      "step": 2960
+    },
+    {
+      "epoch": 2.5201527365294867,
+      "grad_norm": 0.28980498979259595,
+      "learning_rate": 2.897596759391561e-05,
+      "loss": 1.2458,
+      "step": 2970
+    },
+    {
+      "epoch": 2.5286380992787443,
+      "grad_norm": 0.30074882444504475,
+      "learning_rate": 2.8829623983300242e-05,
+      "loss": 1.2498,
+      "step": 2980
+    },
+    {
+      "epoch": 2.537123462028002,
+      "grad_norm": 0.2929721105596463,
+      "learning_rate": 2.868314586750794e-05,
+      "loss": 1.2686,
+      "step": 2990
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "grad_norm": 0.291755235343187,
+      "learning_rate": 2.853653839118605e-05,
+      "loss": 1.2456,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "eval_loss": 1.4051239490509033,
+      "eval_runtime": 52.7875,
+      "eval_samples_per_second": 7.218,
+      "eval_steps_per_second": 0.909,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5540941875265166,
+      "grad_norm": 0.3056527705148328,
+      "learning_rate": 2.8389806703525383e-05,
+      "loss": 1.2321,
+      "step": 3010
+    },
+    {
+      "epoch": 2.562579550275774,
+      "grad_norm": 0.29756401069688737,
+      "learning_rate": 2.8242955958079303e-05,
+      "loss": 1.2341,
+      "step": 3020
+    },
+    {
+      "epoch": 2.571064913025032,
+      "grad_norm": 0.3077048874608071,
+      "learning_rate": 2.809599131258276e-05,
+      "loss": 1.2475,
+      "step": 3030
+    },
+    {
+      "epoch": 2.5795502757742894,
+      "grad_norm": 0.31006088313098146,
+      "learning_rate": 2.7948917928771158e-05,
+      "loss": 1.2381,
+      "step": 3040
+    },
+    {
+      "epoch": 2.588035638523547,
+      "grad_norm": 0.3086227102652305,
+      "learning_rate": 2.7801740972199014e-05,
+      "loss": 1.2386,
+      "step": 3050
+    },
+    {
+      "epoch": 2.596521001272804,
+      "grad_norm": 0.2909420805400902,
+      "learning_rate": 2.7654465612058573e-05,
+      "loss": 1.2071,
+      "step": 3060
+    },
+    {
+      "epoch": 2.6050063640220618,
+      "grad_norm": 0.30310956499188235,
+      "learning_rate": 2.7507097020998246e-05,
+      "loss": 1.2206,
+      "step": 3070
+    },
+    {
+      "epoch": 2.6134917267713194,
+      "grad_norm": 0.2873915382033808,
+      "learning_rate": 2.7359640374940904e-05,
+      "loss": 1.2346,
+      "step": 3080
+    },
+    {
+      "epoch": 2.621977089520577,
+      "grad_norm": 0.29404028686651285,
+      "learning_rate": 2.7212100852902133e-05,
+      "loss": 1.2209,
+      "step": 3090
+    },
+    {
+      "epoch": 2.6304624522698346,
+      "grad_norm": 0.2967558623710032,
+      "learning_rate": 2.7064483636808313e-05,
+      "loss": 1.2471,
+      "step": 3100
+    },
+    {
+      "epoch": 2.638947815019092,
+      "grad_norm": 0.28348844201193973,
+      "learning_rate": 2.6916793911314593e-05,
+      "loss": 1.2271,
+      "step": 3110
+    },
+    {
+      "epoch": 2.6474331777683497,
+      "grad_norm": 0.30908939180701456,
+      "learning_rate": 2.6769036863622842e-05,
+      "loss": 1.2348,
+      "step": 3120
+    },
+    {
+      "epoch": 2.6559185405176073,
+      "grad_norm": 0.296064586506253,
+      "learning_rate": 2.6621217683299437e-05,
+      "loss": 1.2118,
+      "step": 3130
+    },
+    {
+      "epoch": 2.6644039032668645,
+      "grad_norm": 0.29223118946191284,
+      "learning_rate": 2.647334156209299e-05,
+      "loss": 1.2368,
+      "step": 3140
+    },
+    {
+      "epoch": 2.672889266016122,
+      "grad_norm": 0.2974562276968823,
+      "learning_rate": 2.6325413693752004e-05,
+      "loss": 1.2392,
+      "step": 3150
+    },
+    {
+      "epoch": 2.6813746287653797,
+      "grad_norm": 0.30862646184519243,
+      "learning_rate": 2.6177439273842463e-05,
+      "loss": 1.244,
+      "step": 3160
+    },
+    {
+      "epoch": 2.6898599915146373,
+      "grad_norm": 0.2958164221091078,
+      "learning_rate": 2.602942349956536e-05,
+      "loss": 1.2377,
+      "step": 3170
+    },
+    {
+      "epoch": 2.698345354263895,
+      "grad_norm": 0.2941370782364945,
+      "learning_rate": 2.5881371569574125e-05,
+      "loss": 1.2296,
+      "step": 3180
+    },
+    {
+      "epoch": 2.7068307170131525,
+      "grad_norm": 0.2949825785995608,
+      "learning_rate": 2.5733288683792084e-05,
+      "loss": 1.2292,
+      "step": 3190
+    },
+    {
+      "epoch": 2.7153160797624096,
+      "grad_norm": 0.3020705479686205,
+      "learning_rate": 2.558518004322979e-05,
+      "loss": 1.2371,
+      "step": 3200
+    },
+    {
+      "epoch": 2.7238014425116672,
+      "grad_norm": 0.3134012668403533,
+      "learning_rate": 2.5437050849802356e-05,
+      "loss": 1.2257,
+      "step": 3210
+    },
+    {
+      "epoch": 2.732286805260925,
+      "grad_norm": 0.320012175903156,
+      "learning_rate": 2.528890630614677e-05,
+      "loss": 1.215,
+      "step": 3220
+    },
+    {
+      "epoch": 2.7407721680101824,
+      "grad_norm": 0.2936053513063533,
+      "learning_rate": 2.514075161543915e-05,
+      "loss": 1.2364,
+      "step": 3230
+    },
+    {
+      "epoch": 2.74925753075944,
+      "grad_norm": 0.30515854484741317,
+      "learning_rate": 2.499259198121201e-05,
+      "loss": 1.2117,
+      "step": 3240
+    },
+    {
+      "epoch": 2.7577428935086976,
+      "grad_norm": 0.29763910785937486,
+      "learning_rate": 2.484443260717147e-05,
+      "loss": 1.2583,
+      "step": 3250
+    },
+    {
+      "epoch": 2.766228256257955,
+      "grad_norm": 0.2975895109826329,
+      "learning_rate": 2.4696278697014538e-05,
+      "loss": 1.2153,
+      "step": 3260
+    },
+    {
+      "epoch": 2.774713619007213,
+      "grad_norm": 0.2778934357454163,
+      "learning_rate": 2.4548135454246306e-05,
+      "loss": 1.2291,
+      "step": 3270
+    },
+    {
+      "epoch": 2.78319898175647,
+      "grad_norm": 0.32172302391314234,
+      "learning_rate": 2.4400008081997196e-05,
+      "loss": 1.214,
+      "step": 3280
+    },
+    {
+      "epoch": 2.7916843445057276,
+      "grad_norm": 0.3008060855751439,
+      "learning_rate": 2.425190178284024e-05,
+      "loss": 1.234,
+      "step": 3290
+    },
+    {
+      "epoch": 2.800169707254985,
+      "grad_norm": 0.3093107834664431,
+      "learning_rate": 2.4103821758608307e-05,
+      "loss": 1.2492,
+      "step": 3300
+    },
+    {
+      "epoch": 2.8086550700042427,
+      "grad_norm": 0.30554454329766617,
+      "learning_rate": 2.3955773210211465e-05,
+      "loss": 1.2401,
+      "step": 3310
+    },
+    {
+      "epoch": 2.8171404327535003,
+      "grad_norm": 0.2997812898388053,
+      "learning_rate": 2.380776133745425e-05,
+      "loss": 1.2089,
+      "step": 3320
+    },
+    {
+      "epoch": 2.8256257955027575,
+      "grad_norm": 0.2962123179493644,
+      "learning_rate": 2.3659791338853066e-05,
+      "loss": 1.2324,
+      "step": 3330
+    },
+    {
+      "epoch": 2.834111158252015,
+      "grad_norm": 0.31328021490753843,
+      "learning_rate": 2.3511868411453623e-05,
+      "loss": 1.214,
+      "step": 3340
+    },
+    {
+      "epoch": 2.8425965210012727,
+      "grad_norm": 0.30414013736166967,
+      "learning_rate": 2.3363997750648357e-05,
+      "loss": 1.2142,
+      "step": 3350
+    },
+    {
+      "epoch": 2.8510818837505303,
+      "grad_norm": 0.3079683108015467,
+      "learning_rate": 2.3216184549994006e-05,
+      "loss": 1.2137,
+      "step": 3360
+    },
+    {
+      "epoch": 2.859567246499788,
+      "grad_norm": 0.31165537422962203,
+      "learning_rate": 2.3068434001029173e-05,
+      "loss": 1.1915,
+      "step": 3370
+    },
+    {
+      "epoch": 2.8680526092490455,
+      "grad_norm": 0.3190451157121095,
+      "learning_rate": 2.2920751293091948e-05,
+      "loss": 1.2193,
+      "step": 3380
+    },
+    {
+      "epoch": 2.876537971998303,
+      "grad_norm": 0.30247955079343214,
+      "learning_rate": 2.277314161313774e-05,
+      "loss": 1.2253,
+      "step": 3390
+    },
+    {
+      "epoch": 2.8850233347475607,
+      "grad_norm": 0.2936629891547958,
+      "learning_rate": 2.262561014555703e-05,
+      "loss": 1.2136,
+      "step": 3400
+    },
+    {
+      "epoch": 2.893508697496818,
+      "grad_norm": 0.3001872886250926,
+      "learning_rate": 2.2478162071993298e-05,
+      "loss": 1.2061,
+      "step": 3410
+    },
+    {
+      "epoch": 2.9019940602460754,
+      "grad_norm": 0.3111993397471538,
+      "learning_rate": 2.233080257116103e-05,
+      "loss": 1.2193,
+      "step": 3420
+    },
+    {
+      "epoch": 2.910479422995333,
+      "grad_norm": 0.3119972602479391,
+      "learning_rate": 2.2183536818663856e-05,
+      "loss": 1.2125,
+      "step": 3430
+    },
+    {
+      "epoch": 2.9189647857445906,
+      "grad_norm": 0.30787460159489605,
+      "learning_rate": 2.2036369986812713e-05,
+      "loss": 1.215,
+      "step": 3440
+    },
+    {
+      "epoch": 2.927450148493848,
+      "grad_norm": 0.3175794611301324,
+      "learning_rate": 2.1889307244444252e-05,
+      "loss": 1.2202,
+      "step": 3450
+    },
+    {
+      "epoch": 2.9359355112431054,
+      "grad_norm": 0.2956361362699715,
+      "learning_rate": 2.1742353756739247e-05,
+      "loss": 1.2125,
+      "step": 3460
+    },
+    {
+      "epoch": 2.944420873992363,
+      "grad_norm": 0.33942331815482624,
+      "learning_rate": 2.1595514685041205e-05,
+      "loss": 1.2173,
+      "step": 3470
+    },
+    {
+      "epoch": 2.9529062367416206,
+      "grad_norm": 0.332180923100718,
+      "learning_rate": 2.144879518667507e-05,
+      "loss": 1.2266,
+      "step": 3480
+    },
+    {
+      "epoch": 2.961391599490878,
+      "grad_norm": 0.3157083099981729,
+      "learning_rate": 2.1302200414766123e-05,
+      "loss": 1.2154,
+      "step": 3490
+    },
+    {
+      "epoch": 2.9698769622401358,
+      "grad_norm": 0.3149298205272042,
+      "learning_rate": 2.1155735518058914e-05,
+      "loss": 1.2232,
+      "step": 3500
+    },
+    {
+      "epoch": 2.9783623249893934,
+      "grad_norm": 0.3132779789709915,
+      "learning_rate": 2.100940564073653e-05,
+      "loss": 1.2299,
+      "step": 3510
+    },
+    {
+      "epoch": 2.986847687738651,
+      "grad_norm": 0.29387121758746726,
+      "learning_rate": 2.086321592223984e-05,
+      "loss": 1.2219,
+      "step": 3520
+    },
+    {
+      "epoch": 2.9953330504879085,
+      "grad_norm": 0.2954282506485496,
+      "learning_rate": 2.0717171497087014e-05,
+      "loss": 1.2321,
+      "step": 3530
+    },
+    {
+      "epoch": 3.0038184132371657,
+      "grad_norm": 0.2991107469221935,
+      "learning_rate": 2.057127749469321e-05,
+      "loss": 1.2387,
+      "step": 3540
+    },
+    {
+      "epoch": 3.0123037759864233,
+      "grad_norm": 0.3135732189536929,
+      "learning_rate": 2.042553903919036e-05,
+      "loss": 1.1058,
+      "step": 3550
+    },
+    {
+      "epoch": 3.020789138735681,
+      "grad_norm": 0.3179199807851635,
+      "learning_rate": 2.0279961249247274e-05,
+      "loss": 1.0677,
+      "step": 3560
+    },
+    {
+      "epoch": 3.0292745014849385,
+      "grad_norm": 0.3217398418422315,
+      "learning_rate": 2.0134549237889765e-05,
+      "loss": 1.0978,
+      "step": 3570
+    },
+    {
+      "epoch": 3.037759864234196,
+      "grad_norm": 0.32343352041544976,
+      "learning_rate": 1.9989308112321164e-05,
+      "loss": 1.0791,
+      "step": 3580
+    },
+    {
+      "epoch": 3.0462452269834537,
+      "grad_norm": 0.32321267638520695,
+      "learning_rate": 1.9844242973742886e-05,
+      "loss": 1.0991,
+      "step": 3590
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "grad_norm": 0.321242777420917,
+      "learning_rate": 1.9699358917175297e-05,
+      "loss": 1.104,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "eval_loss": 1.3918192386627197,
+      "eval_runtime": 52.2671,
+      "eval_samples_per_second": 7.289,
+      "eval_steps_per_second": 0.918,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0632159524819684,
+      "grad_norm": 0.3313441880623986,
+      "learning_rate": 1.9554661031278712e-05,
+      "loss": 1.081,
+      "step": 3610
+    },
+    {
+      "epoch": 3.071701315231226,
+      "grad_norm": 0.3866269219140372,
+      "learning_rate": 1.9410154398174742e-05,
+      "loss": 1.0826,
+      "step": 3620
+    },
+    {
+      "epoch": 3.0801866779804836,
+      "grad_norm": 0.32719602353062216,
+      "learning_rate": 1.9265844093267728e-05,
+      "loss": 1.0934,
+      "step": 3630
+    },
+    {
+      "epoch": 3.088672040729741,
+      "grad_norm": 0.3301678171750988,
+      "learning_rate": 1.9121735185066537e-05,
+      "loss": 1.1047,
+      "step": 3640
+    },
+    {
+      "epoch": 3.097157403478999,
+      "grad_norm": 0.35644796539197005,
+      "learning_rate": 1.8977832735006522e-05,
+      "loss": 1.0994,
+      "step": 3650
+    },
+    {
+      "epoch": 3.1056427662282564,
+      "grad_norm": 0.33733028940251475,
+      "learning_rate": 1.8834141797271742e-05,
+      "loss": 1.0972,
+      "step": 3660
+    },
+    {
+      "epoch": 3.114128128977514,
+      "grad_norm": 0.32548774712269085,
+      "learning_rate": 1.8690667418617462e-05,
+      "loss": 1.1046,
+      "step": 3670
+    },
+    {
+      "epoch": 3.122613491726771,
+      "grad_norm": 0.3117109384467469,
+      "learning_rate": 1.854741463819291e-05,
+      "loss": 1.0791,
+      "step": 3680
+    },
+    {
+      "epoch": 3.1310988544760288,
+      "grad_norm": 0.3240987091142989,
+      "learning_rate": 1.8404388487364242e-05,
+      "loss": 1.0824,
+      "step": 3690
+    },
+    {
+      "epoch": 3.1395842172252864,
+      "grad_norm": 0.3346401099182515,
+      "learning_rate": 1.8261593989537895e-05,
+      "loss": 1.0753,
+      "step": 3700
+    },
+    {
+      "epoch": 3.148069579974544,
+      "grad_norm": 0.3270030472928521,
+      "learning_rate": 1.81190361599841e-05,
+      "loss": 1.0934,
+      "step": 3710
+    },
+    {
+      "epoch": 3.1565549427238015,
+      "grad_norm": 0.34129438803355183,
+      "learning_rate": 1.797672000566077e-05,
+      "loss": 1.0766,
+      "step": 3720
+    },
+    {
+      "epoch": 3.165040305473059,
+      "grad_norm": 0.37057591586243926,
+      "learning_rate": 1.783465052503762e-05,
+      "loss": 1.1049,
+      "step": 3730
+    },
+    {
+      "epoch": 3.1735256682223163,
+      "grad_norm": 0.3331237337555744,
+      "learning_rate": 1.769283270792065e-05,
+      "loss": 1.0876,
+      "step": 3740
+    },
+    {
+      "epoch": 3.182011030971574,
+      "grad_norm": 0.31904083541369294,
+      "learning_rate": 1.7551271535276792e-05,
+      "loss": 1.1206,
+      "step": 3750
+    },
+    {
+      "epoch": 3.1904963937208315,
+      "grad_norm": 0.35953020953263576,
+      "learning_rate": 1.74099719790591e-05,
+      "loss": 1.0736,
+      "step": 3760
+    },
+    {
+      "epoch": 3.198981756470089,
+      "grad_norm": 0.33595544857573634,
+      "learning_rate": 1.7268939002032035e-05,
+      "loss": 1.0969,
+      "step": 3770
+    },
+    {
+      "epoch": 3.2074671192193467,
+      "grad_norm": 0.34180259956288195,
+      "learning_rate": 1.7128177557597185e-05,
+      "loss": 1.0972,
+      "step": 3780
+    },
+    {
+      "epoch": 3.2159524819686043,
+      "grad_norm": 0.3268247890892541,
+      "learning_rate": 1.6987692589619304e-05,
+      "loss": 1.0737,
+      "step": 3790
+    },
+    {
+      "epoch": 3.224437844717862,
+      "grad_norm": 0.3207792660130559,
+      "learning_rate": 1.6847489032252627e-05,
+      "loss": 1.0797,
+      "step": 3800
+    },
+    {
+      "epoch": 3.232923207467119,
+      "grad_norm": 0.3467270107463577,
+      "learning_rate": 1.6707571809767644e-05,
+      "loss": 1.1024,
+      "step": 3810
+    },
+    {
+      "epoch": 3.2414085702163766,
+      "grad_norm": 0.35579083917156773,
+      "learning_rate": 1.656794583637807e-05,
+      "loss": 1.1026,
+      "step": 3820
+    },
+    {
+      "epoch": 3.2498939329656342,
+      "grad_norm": 0.3339161504484877,
+      "learning_rate": 1.6428616016068304e-05,
+      "loss": 1.0866,
+      "step": 3830
+    },
+    {
+      "epoch": 3.258379295714892,
+      "grad_norm": 0.34920082505086103,
+      "learning_rate": 1.628958724242117e-05,
+      "loss": 1.0789,
+      "step": 3840
+    },
+    {
+      "epoch": 3.2668646584641494,
+      "grad_norm": 0.34847114740056345,
+      "learning_rate": 1.615086439844604e-05,
+      "loss": 1.0849,
+      "step": 3850
+    },
+    {
+      "epoch": 3.275350021213407,
+      "grad_norm": 0.3164317420955498,
+      "learning_rate": 1.601245235640733e-05,
+      "loss": 1.0984,
+      "step": 3860
+    },
+    {
+      "epoch": 3.283835383962664,
+      "grad_norm": 0.32036510223283066,
+      "learning_rate": 1.5874355977653392e-05,
+      "loss": 1.1122,
+      "step": 3870
+    },
+    {
+      "epoch": 3.2923207467119218,
+      "grad_norm": 0.3618051784441363,
+      "learning_rate": 1.5736580112445738e-05,
+      "loss": 1.0942,
+      "step": 3880
+    },
+    {
+      "epoch": 3.3008061094611794,
+      "grad_norm": 0.3376776049321294,
+      "learning_rate": 1.559912959978872e-05,
+      "loss": 1.0898,
+      "step": 3890
+    },
+    {
+      "epoch": 3.309291472210437,
+      "grad_norm": 0.3402374603685398,
+      "learning_rate": 1.546200926725958e-05,
+      "loss": 1.1061,
+      "step": 3900
+    },
+    {
+      "epoch": 3.3177768349596946,
+      "grad_norm": 0.32072793842909514,
+      "learning_rate": 1.5325223930838838e-05,
+      "loss": 1.0995,
+      "step": 3910
+    },
+    {
+      "epoch": 3.326262197708952,
+      "grad_norm": 0.336633795191635,
+      "learning_rate": 1.518877839474122e-05,
+      "loss": 1.0971,
+      "step": 3920
+    },
+    {
+      "epoch": 3.3347475604582097,
+      "grad_norm": 0.36554031780441654,
+      "learning_rate": 1.5052677451246877e-05,
+      "loss": 1.0867,
+      "step": 3930
+    },
+    {
+      "epoch": 3.3432329232074673,
+      "grad_norm": 0.3569408014521597,
+      "learning_rate": 1.491692588053305e-05,
+      "loss": 1.0992,
+      "step": 3940
+    },
+    {
+      "epoch": 3.3517182859567245,
+      "grad_norm": 0.33844743386504,
+      "learning_rate": 1.4781528450506232e-05,
+      "loss": 1.103,
+      "step": 3950
+    },
+    {
+      "epoch": 3.360203648705982,
+      "grad_norm": 0.31581742738052115,
+      "learning_rate": 1.4646489916634687e-05,
+      "loss": 1.0843,
+      "step": 3960
+    },
+    {
+      "epoch": 3.3686890114552397,
+      "grad_norm": 0.3882340052077705,
+      "learning_rate": 1.4511815021781411e-05,
+      "loss": 1.1146,
+      "step": 3970
+    },
+    {
+      "epoch": 3.3771743742044973,
+      "grad_norm": 0.3309843437274728,
+      "learning_rate": 1.4377508496037567e-05,
+      "loss": 1.0751,
+      "step": 3980
+    },
+    {
+      "epoch": 3.385659736953755,
+      "grad_norm": 0.34050270685357675,
+      "learning_rate": 1.4243575056556355e-05,
+      "loss": 1.0995,
+      "step": 3990
+    },
+    {
+      "epoch": 3.3941450997030125,
+      "grad_norm": 0.3459308872000444,
+      "learning_rate": 1.4110019407387315e-05,
+      "loss": 1.0817,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5890,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3362539022843904.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-4000/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

checkpoint-4500/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-72B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.12.0

checkpoint-4500/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-4500/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-4500/latest

@@ -0,0 +1 @@
+global_step4500

checkpoint-4500/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-4500/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-4500/trainer_state.json

@@ -0,0 +1,3239 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.8184132371658888,
+  "eval_steps": 600,
+  "global_step": 4500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00848536274925753,
+      "grad_norm": 0.4898678891363344,
+      "learning_rate": 8.488964346349746e-07,
+      "loss": 1.8056,
+      "step": 10
+    },
+    {
+      "epoch": 0.01697072549851506,
+      "grad_norm": 0.3537473179717183,
+      "learning_rate": 1.6977928692699491e-06,
+      "loss": 1.7621,
+      "step": 20
+    },
+    {
+      "epoch": 0.025456088247772592,
+      "grad_norm": 0.28215953004159977,
+      "learning_rate": 2.546689303904924e-06,
+      "loss": 1.7571,
+      "step": 30
+    },
+    {
+      "epoch": 0.03394145099703012,
+      "grad_norm": 0.27446565146764923,
+      "learning_rate": 3.3955857385398982e-06,
+      "loss": 1.7136,
+      "step": 40
+    },
+    {
+      "epoch": 0.04242681374628765,
+      "grad_norm": 0.17051549768176558,
+      "learning_rate": 4.244482173174873e-06,
+      "loss": 1.6767,
+      "step": 50
+    },
+    {
+      "epoch": 0.050912176495545185,
+      "grad_norm": 0.17763882467320422,
+      "learning_rate": 5.093378607809848e-06,
+      "loss": 1.6371,
+      "step": 60
+    },
+    {
+      "epoch": 0.05939753924480271,
+      "grad_norm": 0.14311462596290048,
+      "learning_rate": 5.942275042444822e-06,
+      "loss": 1.6324,
+      "step": 70
+    },
+    {
+      "epoch": 0.06788290199406025,
+      "grad_norm": 0.1659540846071645,
+      "learning_rate": 6.7911714770797965e-06,
+      "loss": 1.6062,
+      "step": 80
+    },
+    {
+      "epoch": 0.07636826474331777,
+      "grad_norm": 0.20064072815620043,
+      "learning_rate": 7.640067911714771e-06,
+      "loss": 1.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0848536274925753,
+      "grad_norm": 0.2179045681711979,
+      "learning_rate": 8.488964346349745e-06,
+      "loss": 1.5898,
+      "step": 100
+    },
+    {
+      "epoch": 0.09333899024183284,
+      "grad_norm": 0.23866012053128668,
+      "learning_rate": 9.337860780984721e-06,
+      "loss": 1.5924,
+      "step": 110
+    },
+    {
+      "epoch": 0.10182435299109037,
+      "grad_norm": 0.18578051776430282,
+      "learning_rate": 1.0186757215619695e-05,
+      "loss": 1.5877,
+      "step": 120
+    },
+    {
+      "epoch": 0.1103097157403479,
+      "grad_norm": 0.2216509707409362,
+      "learning_rate": 1.103565365025467e-05,
+      "loss": 1.5947,
+      "step": 130
+    },
+    {
+      "epoch": 0.11879507848960542,
+      "grad_norm": 0.20427142255694086,
+      "learning_rate": 1.1884550084889643e-05,
+      "loss": 1.5841,
+      "step": 140
+    },
+    {
+      "epoch": 0.12728044123886295,
+      "grad_norm": 0.1765851415675038,
+      "learning_rate": 1.2733446519524619e-05,
+      "loss": 1.5878,
+      "step": 150
+    },
+    {
+      "epoch": 0.1357658039881205,
+      "grad_norm": 0.1769355117060811,
+      "learning_rate": 1.3582342954159593e-05,
+      "loss": 1.5795,
+      "step": 160
+    },
+    {
+      "epoch": 0.14425116673737803,
+      "grad_norm": 0.1617675663096666,
+      "learning_rate": 1.4431239388794569e-05,
+      "loss": 1.5549,
+      "step": 170
+    },
+    {
+      "epoch": 0.15273652948663555,
+      "grad_norm": 0.17302259072151574,
+      "learning_rate": 1.5280135823429543e-05,
+      "loss": 1.5808,
+      "step": 180
+    },
+    {
+      "epoch": 0.1612218922358931,
+      "grad_norm": 0.16876039012432806,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 1.5676,
+      "step": 190
+    },
+    {
+      "epoch": 0.1697072549851506,
+      "grad_norm": 0.19627360154037596,
+      "learning_rate": 1.697792869269949e-05,
+      "loss": 1.5598,
+      "step": 200
+    },
+    {
+      "epoch": 0.17819261773440814,
+      "grad_norm": 0.16078510362361015,
+      "learning_rate": 1.7826825127334465e-05,
+      "loss": 1.5667,
+      "step": 210
+    },
+    {
+      "epoch": 0.18667798048366568,
+      "grad_norm": 0.16044786518959703,
+      "learning_rate": 1.8675721561969442e-05,
+      "loss": 1.5815,
+      "step": 220
+    },
+    {
+      "epoch": 0.1951633432329232,
+      "grad_norm": 0.15656958873834717,
+      "learning_rate": 1.9524617996604416e-05,
+      "loss": 1.5576,
+      "step": 230
+    },
+    {
+      "epoch": 0.20364870598218074,
+      "grad_norm": 0.1687290471357602,
+      "learning_rate": 2.037351443123939e-05,
+      "loss": 1.5453,
+      "step": 240
+    },
+    {
+      "epoch": 0.21213406873143828,
+      "grad_norm": 0.1519017348276184,
+      "learning_rate": 2.1222410865874364e-05,
+      "loss": 1.5554,
+      "step": 250
+    },
+    {
+      "epoch": 0.2206194314806958,
+      "grad_norm": 0.15761892005160086,
+      "learning_rate": 2.207130730050934e-05,
+      "loss": 1.5494,
+      "step": 260
+    },
+    {
+      "epoch": 0.22910479422995333,
+      "grad_norm": 0.16857088482977495,
+      "learning_rate": 2.2920203735144312e-05,
+      "loss": 1.5794,
+      "step": 270
+    },
+    {
+      "epoch": 0.23759015697921085,
+      "grad_norm": 0.1678705209913503,
+      "learning_rate": 2.3769100169779286e-05,
+      "loss": 1.5373,
+      "step": 280
+    },
+    {
+      "epoch": 0.2460755197284684,
+      "grad_norm": 0.14812649566587394,
+      "learning_rate": 2.461799660441426e-05,
+      "loss": 1.5504,
+      "step": 290
+    },
+    {
+      "epoch": 0.2545608824777259,
+      "grad_norm": 0.17651916734325857,
+      "learning_rate": 2.5466893039049238e-05,
+      "loss": 1.5607,
+      "step": 300
+    },
+    {
+      "epoch": 0.26304624522698344,
+      "grad_norm": 0.14883055338507856,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 1.5311,
+      "step": 310
+    },
+    {
+      "epoch": 0.271531607976241,
+      "grad_norm": 0.15787522753231265,
+      "learning_rate": 2.7164685908319186e-05,
+      "loss": 1.5656,
+      "step": 320
+    },
+    {
+      "epoch": 0.2800169707254985,
+      "grad_norm": 0.1625232940237689,
+      "learning_rate": 2.801358234295416e-05,
+      "loss": 1.5686,
+      "step": 330
+    },
+    {
+      "epoch": 0.28850233347475607,
+      "grad_norm": 0.18505951289343867,
+      "learning_rate": 2.8862478777589137e-05,
+      "loss": 1.5474,
+      "step": 340
+    },
+    {
+      "epoch": 0.29698769622401355,
+      "grad_norm": 0.13785772316349984,
+      "learning_rate": 2.9711375212224108e-05,
+      "loss": 1.5696,
+      "step": 350
+    },
+    {
+      "epoch": 0.3054730589732711,
+      "grad_norm": 0.13531274658248552,
+      "learning_rate": 3.0560271646859086e-05,
+      "loss": 1.5551,
+      "step": 360
+    },
+    {
+      "epoch": 0.31395842172252864,
+      "grad_norm": 0.1366381415368909,
+      "learning_rate": 3.140916808149406e-05,
+      "loss": 1.524,
+      "step": 370
+    },
+    {
+      "epoch": 0.3224437844717862,
+      "grad_norm": 0.14587220569353926,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 1.5515,
+      "step": 380
+    },
+    {
+      "epoch": 0.3309291472210437,
+      "grad_norm": 0.13336349383744864,
+      "learning_rate": 3.310696095076401e-05,
+      "loss": 1.5457,
+      "step": 390
+    },
+    {
+      "epoch": 0.3394145099703012,
+      "grad_norm": 0.1772016947970983,
+      "learning_rate": 3.395585738539898e-05,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 0.34789987271955874,
+      "grad_norm": 0.13819420575084573,
+      "learning_rate": 3.4804753820033956e-05,
+      "loss": 1.5326,
+      "step": 410
+    },
+    {
+      "epoch": 0.3563852354688163,
+      "grad_norm": 0.12729862167862188,
+      "learning_rate": 3.565365025466893e-05,
+      "loss": 1.5387,
+      "step": 420
+    },
+    {
+      "epoch": 0.3648705982180738,
+      "grad_norm": 0.11777082851399363,
+      "learning_rate": 3.6502546689303904e-05,
+      "loss": 1.5587,
+      "step": 430
+    },
+    {
+      "epoch": 0.37335596096733137,
+      "grad_norm": 0.15372268131323022,
+      "learning_rate": 3.7351443123938885e-05,
+      "loss": 1.5362,
+      "step": 440
+    },
+    {
+      "epoch": 0.3818413237165889,
+      "grad_norm": 0.12616185572252248,
+      "learning_rate": 3.820033955857386e-05,
+      "loss": 1.5548,
+      "step": 450
+    },
+    {
+      "epoch": 0.3903266864658464,
+      "grad_norm": 0.1311200786303391,
+      "learning_rate": 3.904923599320883e-05,
+      "loss": 1.5409,
+      "step": 460
+    },
+    {
+      "epoch": 0.39881204921510394,
+      "grad_norm": 0.1707919112561785,
+      "learning_rate": 3.989813242784381e-05,
+      "loss": 1.5509,
+      "step": 470
+    },
+    {
+      "epoch": 0.4072974119643615,
+      "grad_norm": 0.14660149264284913,
+      "learning_rate": 4.074702886247878e-05,
+      "loss": 1.5433,
+      "step": 480
+    },
+    {
+      "epoch": 0.415782774713619,
+      "grad_norm": 0.12478895483834351,
+      "learning_rate": 4.1595925297113755e-05,
+      "loss": 1.5382,
+      "step": 490
+    },
+    {
+      "epoch": 0.42426813746287656,
+      "grad_norm": 0.12327957445795817,
+      "learning_rate": 4.244482173174873e-05,
+      "loss": 1.5515,
+      "step": 500
+    },
+    {
+      "epoch": 0.43275350021213405,
+      "grad_norm": 0.12922777738650987,
+      "learning_rate": 4.32937181663837e-05,
+      "loss": 1.5688,
+      "step": 510
+    },
+    {
+      "epoch": 0.4412388629613916,
+      "grad_norm": 0.12486802189783415,
+      "learning_rate": 4.414261460101868e-05,
+      "loss": 1.5452,
+      "step": 520
+    },
+    {
+      "epoch": 0.44972422571064913,
+      "grad_norm": 0.1360610874577123,
+      "learning_rate": 4.499151103565366e-05,
+      "loss": 1.5493,
+      "step": 530
+    },
+    {
+      "epoch": 0.45820958845990667,
+      "grad_norm": 0.1884897685356775,
+      "learning_rate": 4.5840407470288625e-05,
+      "loss": 1.5511,
+      "step": 540
+    },
+    {
+      "epoch": 0.4666949512091642,
+      "grad_norm": 0.12446302384809525,
+      "learning_rate": 4.6689303904923606e-05,
+      "loss": 1.5458,
+      "step": 550
+    },
+    {
+      "epoch": 0.4751803139584217,
+      "grad_norm": 0.13169591804768588,
+      "learning_rate": 4.753820033955857e-05,
+      "loss": 1.5569,
+      "step": 560
+    },
+    {
+      "epoch": 0.48366567670767924,
+      "grad_norm": 0.1343809247449631,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 1.5408,
+      "step": 570
+    },
+    {
+      "epoch": 0.4921510394569368,
+      "grad_norm": 0.14024589853602,
+      "learning_rate": 4.923599320882852e-05,
+      "loss": 1.5487,
+      "step": 580
+    },
+    {
+      "epoch": 0.5006364022061943,
+      "grad_norm": 0.16240429253875313,
+      "learning_rate": 4.999999560970061e-05,
+      "loss": 1.5488,
+      "step": 590
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "grad_norm": 0.12575424857894482,
+      "learning_rate": 4.999946877563971e-05,
+      "loss": 1.532,
+      "step": 600
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "eval_loss": 1.519254446029663,
+      "eval_runtime": 53.3242,
+      "eval_samples_per_second": 7.145,
+      "eval_steps_per_second": 0.9,
+      "step": 600
+    },
+    {
+      "epoch": 0.5176071277047094,
+      "grad_norm": 0.18688482756329736,
+      "learning_rate": 4.999806390290309e-05,
+      "loss": 1.5544,
+      "step": 610
+    },
+    {
+      "epoch": 0.5260924904539669,
+      "grad_norm": 0.12425469431830571,
+      "learning_rate": 4.999578104083307e-05,
+      "loss": 1.5443,
+      "step": 620
+    },
+    {
+      "epoch": 0.5345778532032245,
+      "grad_norm": 0.1299027485420099,
+      "learning_rate": 4.999262026960902e-05,
+      "loss": 1.5569,
+      "step": 630
+    },
+    {
+      "epoch": 0.543063215952482,
+      "grad_norm": 0.11441754852508934,
+      "learning_rate": 4.998858170024449e-05,
+      "loss": 1.5316,
+      "step": 640
+    },
+    {
+      "epoch": 0.5515485787017395,
+      "grad_norm": 0.14888547248976478,
+      "learning_rate": 4.998366547458326e-05,
+      "loss": 1.5177,
+      "step": 650
+    },
+    {
+      "epoch": 0.560033941450997,
+      "grad_norm": 0.14859292774768867,
+      "learning_rate": 4.997787176529449e-05,
+      "loss": 1.5394,
+      "step": 660
+    },
+    {
+      "epoch": 0.5685193042002545,
+      "grad_norm": 0.12499154376539734,
+      "learning_rate": 4.997120077586651e-05,
+      "loss": 1.5554,
+      "step": 670
+    },
+    {
+      "epoch": 0.5770046669495121,
+      "grad_norm": 0.1218974898058821,
+      "learning_rate": 4.9963652740599774e-05,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.5854900296987696,
+      "grad_norm": 0.1273110498715124,
+      "learning_rate": 4.995522792459859e-05,
+      "loss": 1.5349,
+      "step": 690
+    },
+    {
+      "epoch": 0.5939753924480271,
+      "grad_norm": 0.12115412881719101,
+      "learning_rate": 4.994592662376183e-05,
+      "loss": 1.5419,
+      "step": 700
+    },
+    {
+      "epoch": 0.6024607551972847,
+      "grad_norm": 0.14855096330233286,
+      "learning_rate": 4.99357491647725e-05,
+      "loss": 1.513,
+      "step": 710
+    },
+    {
+      "epoch": 0.6109461179465422,
+      "grad_norm": 0.11407988659327956,
+      "learning_rate": 4.992469590508628e-05,
+      "loss": 1.5243,
+      "step": 720
+    },
+    {
+      "epoch": 0.6194314806957998,
+      "grad_norm": 0.1197712643781127,
+      "learning_rate": 4.9912767232919035e-05,
+      "loss": 1.5177,
+      "step": 730
+    },
+    {
+      "epoch": 0.6279168434450573,
+      "grad_norm": 0.12400515877262065,
+      "learning_rate": 4.9899963567233074e-05,
+      "loss": 1.5619,
+      "step": 740
+    },
+    {
+      "epoch": 0.6364022061943148,
+      "grad_norm": 0.12250385257708406,
+      "learning_rate": 4.988628535772249e-05,
+      "loss": 1.539,
+      "step": 750
+    },
+    {
+      "epoch": 0.6448875689435724,
+      "grad_norm": 0.1262441090496857,
+      "learning_rate": 4.987173308479738e-05,
+      "loss": 1.5195,
+      "step": 760
+    },
+    {
+      "epoch": 0.6533729316928298,
+      "grad_norm": 0.12459694416473029,
+      "learning_rate": 4.985630725956694e-05,
+      "loss": 1.5462,
+      "step": 770
+    },
+    {
+      "epoch": 0.6618582944420874,
+      "grad_norm": 0.12985189006106762,
+      "learning_rate": 4.9840008423821527e-05,
+      "loss": 1.5113,
+      "step": 780
+    },
+    {
+      "epoch": 0.6703436571913449,
+      "grad_norm": 0.12689306141471304,
+      "learning_rate": 4.9822837150013636e-05,
+      "loss": 1.5201,
+      "step": 790
+    },
+    {
+      "epoch": 0.6788290199406024,
+      "grad_norm": 0.15393156370587963,
+      "learning_rate": 4.980479404123778e-05,
+      "loss": 1.5121,
+      "step": 800
+    },
+    {
+      "epoch": 0.68731438268986,
+      "grad_norm": 0.13213701895207608,
+      "learning_rate": 4.978587973120931e-05,
+      "loss": 1.5307,
+      "step": 810
+    },
+    {
+      "epoch": 0.6957997454391175,
+      "grad_norm": 0.11561354931316294,
+      "learning_rate": 4.9766094884242184e-05,
+      "loss": 1.5316,
+      "step": 820
+    },
+    {
+      "epoch": 0.7042851081883751,
+      "grad_norm": 0.12414772399330044,
+      "learning_rate": 4.974544019522559e-05,
+      "loss": 1.5148,
+      "step": 830
+    },
+    {
+      "epoch": 0.7127704709376326,
+      "grad_norm": 0.1171652849153521,
+      "learning_rate": 4.972391638959959e-05,
+      "loss": 1.5096,
+      "step": 840
+    },
+    {
+      "epoch": 0.7212558336868902,
+      "grad_norm": 0.12868937349582316,
+      "learning_rate": 4.9701524223329585e-05,
+      "loss": 1.5282,
+      "step": 850
+    },
+    {
+      "epoch": 0.7297411964361477,
+      "grad_norm": 0.1200015077117309,
+      "learning_rate": 4.967826448287981e-05,
+      "loss": 1.5512,
+      "step": 860
+    },
+    {
+      "epoch": 0.7382265591854051,
+      "grad_norm": 0.12340885660045105,
+      "learning_rate": 4.96541379851857e-05,
+      "loss": 1.5394,
+      "step": 870
+    },
+    {
+      "epoch": 0.7467119219346627,
+      "grad_norm": 0.12976937691467555,
+      "learning_rate": 4.962914557762517e-05,
+      "loss": 1.51,
+      "step": 880
+    },
+    {
+      "epoch": 0.7551972846839202,
+      "grad_norm": 0.11912878476038466,
+      "learning_rate": 4.9603288137988905e-05,
+      "loss": 1.5294,
+      "step": 890
+    },
+    {
+      "epoch": 0.7636826474331778,
+      "grad_norm": 0.1299625480337927,
+      "learning_rate": 4.957656657444947e-05,
+      "loss": 1.507,
+      "step": 900
+    },
+    {
+      "epoch": 0.7721680101824353,
+      "grad_norm": 0.12380144459698468,
+      "learning_rate": 4.954898182552946e-05,
+      "loss": 1.5376,
+      "step": 910
+    },
+    {
+      "epoch": 0.7806533729316928,
+      "grad_norm": 0.13139339643682763,
+      "learning_rate": 4.9520534860068535e-05,
+      "loss": 1.5291,
+      "step": 920
+    },
+    {
+      "epoch": 0.7891387356809504,
+      "grad_norm": 0.13088956203983898,
+      "learning_rate": 4.949122667718935e-05,
+      "loss": 1.5239,
+      "step": 930
+    },
+    {
+      "epoch": 0.7976240984302079,
+      "grad_norm": 0.12586052988453703,
+      "learning_rate": 4.94610583062625e-05,
+      "loss": 1.5525,
+      "step": 940
+    },
+    {
+      "epoch": 0.8061094611794655,
+      "grad_norm": 0.12020996031652877,
+      "learning_rate": 4.943003080687035e-05,
+      "loss": 1.5525,
+      "step": 950
+    },
+    {
+      "epoch": 0.814594823928723,
+      "grad_norm": 0.12866375954060869,
+      "learning_rate": 4.9398145268769856e-05,
+      "loss": 1.5266,
+      "step": 960
+    },
+    {
+      "epoch": 0.8230801866779804,
+      "grad_norm": 0.13166136756817035,
+      "learning_rate": 4.936540281185423e-05,
+      "loss": 1.5041,
+      "step": 970
+    },
+    {
+      "epoch": 0.831565549427238,
+      "grad_norm": 0.12481946698483787,
+      "learning_rate": 4.933180458611364e-05,
+      "loss": 1.5271,
+      "step": 980
+    },
+    {
+      "epoch": 0.8400509121764955,
+      "grad_norm": 0.12264463761209114,
+      "learning_rate": 4.9297351771594844e-05,
+      "loss": 1.5354,
+      "step": 990
+    },
+    {
+      "epoch": 0.8485362749257531,
+      "grad_norm": 0.11985452856537594,
+      "learning_rate": 4.926204557835968e-05,
+      "loss": 1.5167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8570216376750106,
+      "grad_norm": 0.13125396521190327,
+      "learning_rate": 4.9225887246442634e-05,
+      "loss": 1.5282,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8655070004242681,
+      "grad_norm": 0.12730192328072554,
+      "learning_rate": 4.918887804580725e-05,
+      "loss": 1.5089,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8739923631735257,
+      "grad_norm": 0.12724644219344786,
+      "learning_rate": 4.915101927630153e-05,
+      "loss": 1.4964,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8824777259227832,
+      "grad_norm": 0.13578611501833232,
+      "learning_rate": 4.911231226761227e-05,
+      "loss": 1.5189,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8909630886720408,
+      "grad_norm": 0.13577513964986457,
+      "learning_rate": 4.90727583792184e-05,
+      "loss": 1.5149,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8994484514212983,
+      "grad_norm": 0.1269735011676505,
+      "learning_rate": 4.903235900034317e-05,
+      "loss": 1.5066,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9079338141705557,
+      "grad_norm": 0.13250058214235566,
+      "learning_rate": 4.899111554990543e-05,
+      "loss": 1.5129,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9164191769198133,
+      "grad_norm": 0.13130735246433495,
+      "learning_rate": 4.894902947646975e-05,
+      "loss": 1.5156,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9249045396690708,
+      "grad_norm": 0.1273580180253049,
+      "learning_rate": 4.890610225819553e-05,
+      "loss": 1.5324,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9333899024183284,
+      "grad_norm": 0.13155314243939242,
+      "learning_rate": 4.8862335402785136e-05,
+      "loss": 1.5106,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9418752651675859,
+      "grad_norm": 0.13564895211984299,
+      "learning_rate": 4.88177304474309e-05,
+      "loss": 1.5067,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9503606279168434,
+      "grad_norm": 0.12774735587114736,
+      "learning_rate": 4.877228895876115e-05,
+      "loss": 1.5182,
+      "step": 1120
+    },
+    {
+      "epoch": 0.958845990666101,
+      "grad_norm": 0.1307997709537685,
+      "learning_rate": 4.872601253278517e-05,
+      "loss": 1.4969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9673313534153585,
+      "grad_norm": 0.1304794845040634,
+      "learning_rate": 4.867890279483717e-05,
+      "loss": 1.5264,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9758167161646161,
+      "grad_norm": 0.13666141796489684,
+      "learning_rate": 4.8630961399519206e-05,
+      "loss": 1.5467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9843020789138736,
+      "grad_norm": 0.1370278303190263,
+      "learning_rate": 4.8582190030643e-05,
+      "loss": 1.5127,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9927874416631312,
+      "grad_norm": 0.1390936629299565,
+      "learning_rate": 4.8532590401170894e-05,
+      "loss": 1.5058,
+      "step": 1170
+    },
+    {
+      "epoch": 1.0012728044123886,
+      "grad_norm": 0.12934475548108287,
+      "learning_rate": 4.848216425315561e-05,
+      "loss": 1.5202,
+      "step": 1180
+    },
+    {
+      "epoch": 1.0097581671616462,
+      "grad_norm": 0.13898591683370803,
+      "learning_rate": 4.843091335767913e-05,
+      "loss": 1.4563,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "grad_norm": 0.17488231535826249,
+      "learning_rate": 4.837883951479043e-05,
+      "loss": 1.4402,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "eval_loss": 1.4955657720565796,
+      "eval_runtime": 52.424,
+      "eval_samples_per_second": 7.268,
+      "eval_steps_per_second": 0.916,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0267288926601612,
+      "grad_norm": 0.1536036344095855,
+      "learning_rate": 4.832594455344229e-05,
+      "loss": 1.4848,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0352142554094188,
+      "grad_norm": 0.15762414421336599,
+      "learning_rate": 4.827223033142706e-05,
+      "loss": 1.4567,
+      "step": 1220
+    },
+    {
+      "epoch": 1.0436996181586762,
+      "grad_norm": 0.15058229398130366,
+      "learning_rate": 4.8217698735311414e-05,
+      "loss": 1.4672,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0521849809079338,
+      "grad_norm": 0.16010992835678386,
+      "learning_rate": 4.8162351680370044e-05,
+      "loss": 1.4458,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0606703436571914,
+      "grad_norm": 0.16758816000341356,
+      "learning_rate": 4.810619111051847e-05,
+      "loss": 1.4842,
+      "step": 1250
+    },
+    {
+      "epoch": 1.069155706406449,
+      "grad_norm": 0.16559260972674986,
+      "learning_rate": 4.8049218998244696e-05,
+      "loss": 1.4556,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0776410691557063,
+      "grad_norm": 0.17237632034416966,
+      "learning_rate": 4.7991437344539966e-05,
+      "loss": 1.4813,
+      "step": 1270
+    },
+    {
+      "epoch": 1.086126431904964,
+      "grad_norm": 0.17112756741722487,
+      "learning_rate": 4.793284817882845e-05,
+      "loss": 1.4535,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0946117946542215,
+      "grad_norm": 0.16828572707718548,
+      "learning_rate": 4.787345355889604e-05,
+      "loss": 1.4344,
+      "step": 1290
+    },
+    {
+      "epoch": 1.103097157403479,
+      "grad_norm": 0.15709986047041227,
+      "learning_rate": 4.7813255570817985e-05,
+      "loss": 1.4744,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1115825201527365,
+      "grad_norm": 0.16651547128146313,
+      "learning_rate": 4.775225632888568e-05,
+      "loss": 1.4561,
+      "step": 1310
+    },
+    {
+      "epoch": 1.120067882901994,
+      "grad_norm": 0.16750176017515714,
+      "learning_rate": 4.76904579755324e-05,
+      "loss": 1.4616,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1285532456512515,
+      "grad_norm": 0.1608016567554825,
+      "learning_rate": 4.7627862681258037e-05,
+      "loss": 1.4593,
+      "step": 1330
+    },
+    {
+      "epoch": 1.137038608400509,
+      "grad_norm": 0.21390766919038295,
+      "learning_rate": 4.756447264455287e-05,
+      "loss": 1.4484,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1455239711497667,
+      "grad_norm": 0.16826883293172662,
+      "learning_rate": 4.750029009182038e-05,
+      "loss": 1.4703,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1540093338990243,
+      "grad_norm": 0.17431508867079595,
+      "learning_rate": 4.7435317277299e-05,
+      "loss": 1.4701,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1624946966482816,
+      "grad_norm": 0.15973851467570443,
+      "learning_rate": 4.736955648298299e-05,
+      "loss": 1.4503,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1709800593975392,
+      "grad_norm": 0.1887713767970947,
+      "learning_rate": 4.730301001854225e-05,
+      "loss": 1.4624,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1794654221467968,
+      "grad_norm": 0.16898695344997974,
+      "learning_rate": 4.7235680221241216e-05,
+      "loss": 1.4452,
+      "step": 1390
+    },
+    {
+      "epoch": 1.1879507848960542,
+      "grad_norm": 0.20014553287073528,
+      "learning_rate": 4.716756945585681e-05,
+      "loss": 1.4717,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1964361476453118,
+      "grad_norm": 0.17137954325200072,
+      "learning_rate": 4.709868011459528e-05,
+      "loss": 1.4403,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2049215103945694,
+      "grad_norm": 0.17801721751888322,
+      "learning_rate": 4.7029014617008294e-05,
+      "loss": 1.4339,
+      "step": 1420
+    },
+    {
+      "epoch": 1.213406873143827,
+      "grad_norm": 0.17139613676642362,
+      "learning_rate": 4.695857540990789e-05,
+      "loss": 1.4573,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2218922358930844,
+      "grad_norm": 0.16971403514498054,
+      "learning_rate": 4.688736496728058e-05,
+      "loss": 1.4282,
+      "step": 1440
+    },
+    {
+      "epoch": 1.230377598642342,
+      "grad_norm": 0.17200272420880428,
+      "learning_rate": 4.681538579020038e-05,
+      "loss": 1.4434,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2388629613915996,
+      "grad_norm": 0.17208160407432616,
+      "learning_rate": 4.6742640406741106e-05,
+      "loss": 1.45,
+      "step": 1460
+    },
+    {
+      "epoch": 1.247348324140857,
+      "grad_norm": 0.1939626212901777,
+      "learning_rate": 4.666913137188743e-05,
+      "loss": 1.4608,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2558336868901145,
+      "grad_norm": 0.17291794493304186,
+      "learning_rate": 4.6594861267445236e-05,
+      "loss": 1.4671,
+      "step": 1480
+    },
+    {
+      "epoch": 1.2643190496393721,
+      "grad_norm": 0.18219792041638924,
+      "learning_rate": 4.651983270195093e-05,
+      "loss": 1.4262,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2728044123886297,
+      "grad_norm": 0.18086437830489926,
+      "learning_rate": 4.644404831057979e-05,
+      "loss": 1.4455,
+      "step": 1500
+    },
+    {
+      "epoch": 1.281289775137887,
+      "grad_norm": 0.17417619624549402,
+      "learning_rate": 4.636751075505344e-05,
+      "loss": 1.4873,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2897751378871447,
+      "grad_norm": 0.18354282411845188,
+      "learning_rate": 4.629022272354637e-05,
+      "loss": 1.4525,
+      "step": 1520
+    },
+    {
+      "epoch": 1.298260500636402,
+      "grad_norm": 0.17985617345325455,
+      "learning_rate": 4.621218693059149e-05,
+      "loss": 1.4303,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3067458633856597,
+      "grad_norm": 0.1809708317849863,
+      "learning_rate": 4.6133406116984795e-05,
+      "loss": 1.4631,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3152312261349173,
+      "grad_norm": 0.17487374671212322,
+      "learning_rate": 4.6053883049689145e-05,
+      "loss": 1.4482,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3237165888841749,
+      "grad_norm": 0.19912807671077193,
+      "learning_rate": 4.5973620521737036e-05,
+      "loss": 1.4497,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3322019516334322,
+      "grad_norm": 0.17853627546912074,
+      "learning_rate": 4.5892621352132514e-05,
+      "loss": 1.4456,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3406873143826898,
+      "grad_norm": 0.18252596927754394,
+      "learning_rate": 4.581088838575218e-05,
+      "loss": 1.4328,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3491726771319474,
+      "grad_norm": 0.17604951053556211,
+      "learning_rate": 4.572842449324525e-05,
+      "loss": 1.4442,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3576580398812048,
+      "grad_norm": 0.18358942463311748,
+      "learning_rate": 4.564523257093275e-05,
+      "loss": 1.4338,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3661434026304624,
+      "grad_norm": 0.20508703236267142,
+      "learning_rate": 4.5561315540705774e-05,
+      "loss": 1.4445,
+      "step": 1610
+    },
+    {
+      "epoch": 1.37462876537972,
+      "grad_norm": 0.18486352550747187,
+      "learning_rate": 4.547667634992288e-05,
+      "loss": 1.4261,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3831141281289776,
+      "grad_norm": 0.17492766465456316,
+      "learning_rate": 4.539131797130656e-05,
+      "loss": 1.4258,
+      "step": 1630
+    },
+    {
+      "epoch": 1.391599490878235,
+      "grad_norm": 0.19692876587833674,
+      "learning_rate": 4.530524340283881e-05,
+      "loss": 1.4349,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4000848536274926,
+      "grad_norm": 0.19155373430892478,
+      "learning_rate": 4.521845566765589e-05,
+      "loss": 1.4536,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4085702163767502,
+      "grad_norm": 0.18544325977459192,
+      "learning_rate": 4.513095781394208e-05,
+      "loss": 1.4363,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4170555791260075,
+      "grad_norm": 0.177828004720666,
+      "learning_rate": 4.504275291482267e-05,
+      "loss": 1.4595,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4255409418752651,
+      "grad_norm": 0.17855432230356816,
+      "learning_rate": 4.495384406825601e-05,
+      "loss": 1.4211,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4340263046245227,
+      "grad_norm": 0.20232492538380317,
+      "learning_rate": 4.486423439692469e-05,
+      "loss": 1.4189,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4425116673737803,
+      "grad_norm": 0.1975109303350431,
+      "learning_rate": 4.477392704812585e-05,
+      "loss": 1.4565,
+      "step": 1700
+    },
+    {
+      "epoch": 1.4509970301230377,
+      "grad_norm": 0.19619010830399825,
+      "learning_rate": 4.468292519366071e-05,
+      "loss": 1.4382,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4594823928722953,
+      "grad_norm": 0.18168826428246143,
+      "learning_rate": 4.459123202972308e-05,
+      "loss": 1.4471,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4679677556215527,
+      "grad_norm": 0.1923264062362399,
+      "learning_rate": 4.449885077678717e-05,
+      "loss": 1.4153,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4764531183708103,
+      "grad_norm": 0.1907937313040222,
+      "learning_rate": 4.440578467949445e-05,
+      "loss": 1.4432,
+      "step": 1740
+    },
+    {
+      "epoch": 1.4849384811200679,
+      "grad_norm": 0.19107457667767244,
+      "learning_rate": 4.431203700653968e-05,
+      "loss": 1.4285,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4934238438693255,
+      "grad_norm": 0.19847350429107552,
+      "learning_rate": 4.421761105055613e-05,
+      "loss": 1.4383,
+      "step": 1760
+    },
+    {
+      "epoch": 1.501909206618583,
+      "grad_norm": 0.18536475556610216,
+      "learning_rate": 4.4122510127999937e-05,
+      "loss": 1.42,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5103945693678404,
+      "grad_norm": 0.18481023473586697,
+      "learning_rate": 4.4026737579033584e-05,
+      "loss": 1.4384,
+      "step": 1780
+    },
+    {
+      "epoch": 1.518879932117098,
+      "grad_norm": 0.20863867505874642,
+      "learning_rate": 4.393029676740864e-05,
+      "loss": 1.4543,
+      "step": 1790
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "grad_norm": 0.1816036870853105,
+      "learning_rate": 4.3833191080347575e-05,
+      "loss": 1.434,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "eval_loss": 1.4622184038162231,
+      "eval_runtime": 52.4041,
+      "eval_samples_per_second": 7.27,
+      "eval_steps_per_second": 0.916,
+      "step": 1800
+    },
+    {
+      "epoch": 1.535850657615613,
+      "grad_norm": 0.19378252368958881,
+      "learning_rate": 4.3735423928424815e-05,
+      "loss": 1.4275,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5443360203648706,
+      "grad_norm": 0.20453331251433848,
+      "learning_rate": 4.363699874544697e-05,
+      "loss": 1.4203,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5528213831141282,
+      "grad_norm": 0.26684319417219377,
+      "learning_rate": 4.3537918988332156e-05,
+      "loss": 1.4372,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5613067458633858,
+      "grad_norm": 0.25745160303419773,
+      "learning_rate": 4.343818813698868e-05,
+      "loss": 1.4082,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5697921086126432,
+      "grad_norm": 0.19969727996700776,
+      "learning_rate": 4.3337809694192765e-05,
+      "loss": 1.4314,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5782774713619008,
+      "grad_norm": 0.20117210832277968,
+      "learning_rate": 4.3236787185465525e-05,
+      "loss": 1.4293,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5867628341111581,
+      "grad_norm": 0.20173003641028897,
+      "learning_rate": 4.313512415894913e-05,
+      "loss": 1.4406,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5952481968604157,
+      "grad_norm": 0.20304770794371527,
+      "learning_rate": 4.303282418528224e-05,
+      "loss": 1.4286,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6037335596096733,
+      "grad_norm": 0.19126658907738198,
+      "learning_rate": 4.292989085747452e-05,
+      "loss": 1.4184,
+      "step": 1890
+    },
+    {
+      "epoch": 1.612218922358931,
+      "grad_norm": 0.20069554966453027,
+      "learning_rate": 4.282632779078051e-05,
+      "loss": 1.4133,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6207042851081885,
+      "grad_norm": 0.1952881519566686,
+      "learning_rate": 4.2722138622572624e-05,
+      "loss": 1.4432,
+      "step": 1910
+    },
+    {
+      "epoch": 1.629189647857446,
+      "grad_norm": 0.19763704668680288,
+      "learning_rate": 4.261732701221339e-05,
+      "loss": 1.3921,
+      "step": 1920
+    },
+    {
+      "epoch": 1.6376750106067033,
+      "grad_norm": 0.19821464294464497,
+      "learning_rate": 4.2511896640926925e-05,
+      "loss": 1.4454,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6461603733559609,
+      "grad_norm": 0.20456545626297834,
+      "learning_rate": 4.240585121166966e-05,
+      "loss": 1.4147,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6546457361052185,
+      "grad_norm": 0.2119092529186395,
+      "learning_rate": 4.229919444900027e-05,
+      "loss": 1.3969,
+      "step": 1950
+    },
+    {
+      "epoch": 1.663131098854476,
+      "grad_norm": 0.20330157582122357,
+      "learning_rate": 4.2191930098948865e-05,
+      "loss": 1.426,
+      "step": 1960
+    },
+    {
+      "epoch": 1.6716164616037337,
+      "grad_norm": 0.21761164739298738,
+      "learning_rate": 4.2084061928885406e-05,
+      "loss": 1.4246,
+      "step": 1970
+    },
+    {
+      "epoch": 1.680101824352991,
+      "grad_norm": 0.19331588142071401,
+      "learning_rate": 4.197559372738741e-05,
+      "loss": 1.4305,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6885871871022486,
+      "grad_norm": 0.20188460724329996,
+      "learning_rate": 4.186652930410685e-05,
+      "loss": 1.4153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.697072549851506,
+      "grad_norm": 0.20988950033571588,
+      "learning_rate": 4.1756872489636425e-05,
+      "loss": 1.3894,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7055579126007636,
+      "grad_norm": 0.1966475893123187,
+      "learning_rate": 4.1646627135374916e-05,
+      "loss": 1.3962,
+      "step": 2010
+    },
+    {
+      "epoch": 1.7140432753500212,
+      "grad_norm": 0.20785207367991768,
+      "learning_rate": 4.1535797113392004e-05,
+      "loss": 1.4037,
+      "step": 2020
+    },
+    {
+      "epoch": 1.7225286380992788,
+      "grad_norm": 0.2029940281663133,
+      "learning_rate": 4.1424386316292224e-05,
+      "loss": 1.4011,
+      "step": 2030
+    },
+    {
+      "epoch": 1.7310140008485364,
+      "grad_norm": 0.2247844551379277,
+      "learning_rate": 4.131239865707829e-05,
+      "loss": 1.4084,
+      "step": 2040
+    },
+    {
+      "epoch": 1.7394993635977938,
+      "grad_norm": 0.20900441746105022,
+      "learning_rate": 4.11998380690136e-05,
+      "loss": 1.4235,
+      "step": 2050
+    },
+    {
+      "epoch": 1.7479847263470514,
+      "grad_norm": 0.20362408546889926,
+      "learning_rate": 4.108670850548416e-05,
+      "loss": 1.4204,
+      "step": 2060
+    },
+    {
+      "epoch": 1.7564700890963088,
+      "grad_norm": 0.22281567946240438,
+      "learning_rate": 4.097301393985968e-05,
+      "loss": 1.4023,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7649554518455663,
+      "grad_norm": 0.20867113178797225,
+      "learning_rate": 4.085875836535404e-05,
+      "loss": 1.3895,
+      "step": 2080
+    },
+    {
+      "epoch": 1.773440814594824,
+      "grad_norm": 0.22113231886160947,
+      "learning_rate": 4.0743945794885063e-05,
+      "loss": 1.3963,
+      "step": 2090
+    },
+    {
+      "epoch": 1.7819261773440815,
+      "grad_norm": 0.22334563577844263,
+      "learning_rate": 4.062858026093351e-05,
+      "loss": 1.3988,
+      "step": 2100
+    },
+    {
+      "epoch": 1.7904115400933391,
+      "grad_norm": 0.23218581668265403,
+      "learning_rate": 4.051266581540152e-05,
+      "loss": 1.4068,
+      "step": 2110
+    },
+    {
+      "epoch": 1.7988969028425965,
+      "grad_norm": 0.20295589384571033,
+      "learning_rate": 4.0396206529470234e-05,
+      "loss": 1.3883,
+      "step": 2120
+    },
+    {
+      "epoch": 1.8073822655918539,
+      "grad_norm": 0.22861611442392848,
+      "learning_rate": 4.027920649345687e-05,
+      "loss": 1.4043,
+      "step": 2130
+    },
+    {
+      "epoch": 1.8158676283411115,
+      "grad_norm": 0.2083012771089638,
+      "learning_rate": 4.0161669816671e-05,
+      "loss": 1.398,
+      "step": 2140
+    },
+    {
+      "epoch": 1.824352991090369,
+      "grad_norm": 0.21936173231840464,
+      "learning_rate": 4.004360062727028e-05,
+      "loss": 1.4142,
+      "step": 2150
+    },
+    {
+      "epoch": 1.8328383538396267,
+      "grad_norm": 0.21383435796328337,
+      "learning_rate": 3.9925003072115406e-05,
+      "loss": 1.4138,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8413237165888843,
+      "grad_norm": 0.23301608248270392,
+      "learning_rate": 3.9805881316624506e-05,
+      "loss": 1.4195,
+      "step": 2170
+    },
+    {
+      "epoch": 1.8498090793381419,
+      "grad_norm": 0.22424766656883474,
+      "learning_rate": 3.968623954462681e-05,
+      "loss": 1.4011,
+      "step": 2180
+    },
+    {
+      "epoch": 1.8582944420873992,
+      "grad_norm": 0.21286417342881453,
+      "learning_rate": 3.9566081958215734e-05,
+      "loss": 1.409,
+      "step": 2190
+    },
+    {
+      "epoch": 1.8667798048366566,
+      "grad_norm": 0.21944800687444807,
+      "learning_rate": 3.9445412777601284e-05,
+      "loss": 1.3877,
+      "step": 2200
+    },
+    {
+      "epoch": 1.8752651675859142,
+      "grad_norm": 0.23113173625974803,
+      "learning_rate": 3.932423624096181e-05,
+      "loss": 1.4089,
+      "step": 2210
+    },
+    {
+      "epoch": 1.8837505303351718,
+      "grad_norm": 0.2081941699587778,
+      "learning_rate": 3.920255660429517e-05,
+      "loss": 1.4024,
+      "step": 2220
+    },
+    {
+      "epoch": 1.8922358930844294,
+      "grad_norm": 0.2188685806654701,
+      "learning_rate": 3.908037814126927e-05,
+      "loss": 1.3878,
+      "step": 2230
+    },
+    {
+      "epoch": 1.900721255833687,
+      "grad_norm": 0.22761843244757962,
+      "learning_rate": 3.895770514307193e-05,
+      "loss": 1.4004,
+      "step": 2240
+    },
+    {
+      "epoch": 1.9092066185829444,
+      "grad_norm": 0.23309183623120422,
+      "learning_rate": 3.883454191826017e-05,
+      "loss": 1.4188,
+      "step": 2250
+    },
+    {
+      "epoch": 1.917691981332202,
+      "grad_norm": 0.20329785843911802,
+      "learning_rate": 3.871089279260891e-05,
+      "loss": 1.3893,
+      "step": 2260
+    },
+    {
+      "epoch": 1.9261773440814594,
+      "grad_norm": 0.23470973193726366,
+      "learning_rate": 3.8586762108958995e-05,
+      "loss": 1.3974,
+      "step": 2270
+    },
+    {
+      "epoch": 1.934662706830717,
+      "grad_norm": 0.22779136837044714,
+      "learning_rate": 3.8462154227064725e-05,
+      "loss": 1.4115,
+      "step": 2280
+    },
+    {
+      "epoch": 1.9431480695799745,
+      "grad_norm": 0.22338952315651892,
+      "learning_rate": 3.833707352344068e-05,
+      "loss": 1.3873,
+      "step": 2290
+    },
+    {
+      "epoch": 1.9516334323292321,
+      "grad_norm": 0.23069304025882129,
+      "learning_rate": 3.821152439120801e-05,
+      "loss": 1.3944,
+      "step": 2300
+    },
+    {
+      "epoch": 1.9601187950784897,
+      "grad_norm": 0.23590596270163203,
+      "learning_rate": 3.808551123994018e-05,
+      "loss": 1.3857,
+      "step": 2310
+    },
+    {
+      "epoch": 1.9686041578277471,
+      "grad_norm": 0.22545661808214923,
+      "learning_rate": 3.795903849550805e-05,
+      "loss": 1.3628,
+      "step": 2320
+    },
+    {
+      "epoch": 1.9770895205770047,
+      "grad_norm": 0.2450769875954842,
+      "learning_rate": 3.7832110599924455e-05,
+      "loss": 1.4079,
+      "step": 2330
+    },
+    {
+      "epoch": 1.985574883326262,
+      "grad_norm": 0.22931499326784313,
+      "learning_rate": 3.7704732011188166e-05,
+      "loss": 1.379,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9940602460755197,
+      "grad_norm": 0.22417244507397657,
+      "learning_rate": 3.7576907203127346e-05,
+      "loss": 1.4035,
+      "step": 2350
+    },
+    {
+      "epoch": 2.0025456088247773,
+      "grad_norm": 0.24496197221575314,
+      "learning_rate": 3.7448640665242406e-05,
+      "loss": 1.442,
+      "step": 2360
+    },
+    {
+      "epoch": 2.011030971574035,
+      "grad_norm": 0.2532740296990078,
+      "learning_rate": 3.73199369025483e-05,
+      "loss": 1.2672,
+      "step": 2370
+    },
+    {
+      "epoch": 2.0195163343232925,
+      "grad_norm": 0.2890155987968593,
+      "learning_rate": 3.7190800435416355e-05,
+      "loss": 1.246,
+      "step": 2380
+    },
+    {
+      "epoch": 2.02800169707255,
+      "grad_norm": 0.2541972565696406,
+      "learning_rate": 3.706123579941545e-05,
+      "loss": 1.2603,
+      "step": 2390
+    },
+    {
+      "epoch": 2.036487059821807,
+      "grad_norm": 0.2530140862527023,
+      "learning_rate": 3.693124754515272e-05,
+      "loss": 1.2638,
+      "step": 2400
+    },
+    {
+      "epoch": 2.036487059821807,
+      "eval_loss": 1.435962438583374,
+      "eval_runtime": 52.582,
+      "eval_samples_per_second": 7.246,
+      "eval_steps_per_second": 0.913,
+      "step": 2400
+    },
+    {
+      "epoch": 2.044972422571065,
+      "grad_norm": 0.25100458343337734,
+      "learning_rate": 3.680084023811377e-05,
+      "loss": 1.2711,
+      "step": 2410
+    },
+    {
+      "epoch": 2.0534577853203224,
+      "grad_norm": 0.2695727673292618,
+      "learning_rate": 3.66700184585023e-05,
+      "loss": 1.2578,
+      "step": 2420
+    },
+    {
+      "epoch": 2.06194314806958,
+      "grad_norm": 0.2605068415443213,
+      "learning_rate": 3.6538786801079226e-05,
+      "loss": 1.2506,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0704285108188376,
+      "grad_norm": 0.27415607207865045,
+      "learning_rate": 3.64071498750013e-05,
+      "loss": 1.2852,
+      "step": 2440
+    },
+    {
+      "epoch": 2.078913873568095,
+      "grad_norm": 0.2688900338206285,
+      "learning_rate": 3.627511230365928e-05,
+      "loss": 1.2695,
+      "step": 2450
+    },
+    {
+      "epoch": 2.0873992363173524,
+      "grad_norm": 0.2750825805336503,
+      "learning_rate": 3.614267872451546e-05,
+      "loss": 1.2643,
+      "step": 2460
+    },
+    {
+      "epoch": 2.09588459906661,
+      "grad_norm": 0.2659269066581903,
+      "learning_rate": 3.600985378894086e-05,
+      "loss": 1.2868,
+      "step": 2470
+    },
+    {
+      "epoch": 2.1043699618158676,
+      "grad_norm": 0.24411151291321526,
+      "learning_rate": 3.587664216205183e-05,
+      "loss": 1.2571,
+      "step": 2480
+    },
+    {
+      "epoch": 2.112855324565125,
+      "grad_norm": 0.2574194755634052,
+      "learning_rate": 3.574304852254621e-05,
+      "loss": 1.2769,
+      "step": 2490
+    },
+    {
+      "epoch": 2.1213406873143827,
+      "grad_norm": 0.2894545074998905,
+      "learning_rate": 3.5609077562538997e-05,
+      "loss": 1.2469,
+      "step": 2500
+    },
+    {
+      "epoch": 2.1298260500636403,
+      "grad_norm": 0.2828176429904294,
+      "learning_rate": 3.547473398739754e-05,
+      "loss": 1.2527,
+      "step": 2510
+    },
+    {
+      "epoch": 2.138311412812898,
+      "grad_norm": 0.25886029771650565,
+      "learning_rate": 3.5340022515576294e-05,
+      "loss": 1.2578,
+      "step": 2520
+    },
+    {
+      "epoch": 2.146796775562155,
+      "grad_norm": 0.2783799371621383,
+      "learning_rate": 3.52049478784511e-05,
+      "loss": 1.2489,
+      "step": 2530
+    },
+    {
+      "epoch": 2.1552821383114127,
+      "grad_norm": 0.2753116113218978,
+      "learning_rate": 3.506951482015297e-05,
+      "loss": 1.275,
+      "step": 2540
+    },
+    {
+      "epoch": 2.1637675010606703,
+      "grad_norm": 0.28115792079727675,
+      "learning_rate": 3.493372809740152e-05,
+      "loss": 1.2554,
+      "step": 2550
+    },
+    {
+      "epoch": 2.172252863809928,
+      "grad_norm": 0.27954425325951715,
+      "learning_rate": 3.479759247933785e-05,
+      "loss": 1.2618,
+      "step": 2560
+    },
+    {
+      "epoch": 2.1807382265591855,
+      "grad_norm": 0.27555174232347995,
+      "learning_rate": 3.466111274735707e-05,
+      "loss": 1.2598,
+      "step": 2570
+    },
+    {
+      "epoch": 2.189223589308443,
+      "grad_norm": 0.27280827991301104,
+      "learning_rate": 3.452429369494037e-05,
+      "loss": 1.262,
+      "step": 2580
+    },
+    {
+      "epoch": 2.1977089520577007,
+      "grad_norm": 0.2749685805551003,
+      "learning_rate": 3.438714012748664e-05,
+      "loss": 1.2683,
+      "step": 2590
+    },
+    {
+      "epoch": 2.206194314806958,
+      "grad_norm": 0.2780594302788235,
+      "learning_rate": 3.424965686214371e-05,
+      "loss": 1.2462,
+      "step": 2600
+    },
+    {
+      "epoch": 2.2146796775562154,
+      "grad_norm": 0.2942257416636676,
+      "learning_rate": 3.411184872763915e-05,
+      "loss": 1.2581,
+      "step": 2610
+    },
+    {
+      "epoch": 2.223165040305473,
+      "grad_norm": 0.27000377333423803,
+      "learning_rate": 3.39737205641107e-05,
+      "loss": 1.2412,
+      "step": 2620
+    },
+    {
+      "epoch": 2.2316504030547306,
+      "grad_norm": 0.28187507810449336,
+      "learning_rate": 3.383527722293622e-05,
+      "loss": 1.2659,
+      "step": 2630
+    },
+    {
+      "epoch": 2.240135765803988,
+      "grad_norm": 0.2736213940552268,
+      "learning_rate": 3.369652356656336e-05,
+      "loss": 1.2553,
+      "step": 2640
+    },
+    {
+      "epoch": 2.248621128553246,
+      "grad_norm": 0.29698834543438446,
+      "learning_rate": 3.355746446833873e-05,
+      "loss": 1.2714,
+      "step": 2650
+    },
+    {
+      "epoch": 2.257106491302503,
+      "grad_norm": 0.2875128112484735,
+      "learning_rate": 3.3418104812336786e-05,
+      "loss": 1.2508,
+      "step": 2660
+    },
+    {
+      "epoch": 2.2655918540517606,
+      "grad_norm": 0.3016647299373059,
+      "learning_rate": 3.327844949318824e-05,
+      "loss": 1.2451,
+      "step": 2670
+    },
+    {
+      "epoch": 2.274077216801018,
+      "grad_norm": 0.27371321581702696,
+      "learning_rate": 3.3138503415908176e-05,
+      "loss": 1.2467,
+      "step": 2680
+    },
+    {
+      "epoch": 2.2825625795502757,
+      "grad_norm": 0.28374547760120017,
+      "learning_rate": 3.299827149572376e-05,
+      "loss": 1.2452,
+      "step": 2690
+    },
+    {
+      "epoch": 2.2910479422995333,
+      "grad_norm": 0.2805999278165284,
+      "learning_rate": 3.285775865790166e-05,
+      "loss": 1.2595,
+      "step": 2700
+    },
+    {
+      "epoch": 2.299533305048791,
+      "grad_norm": 0.2758019804125597,
+      "learning_rate": 3.271696983757496e-05,
+      "loss": 1.2583,
+      "step": 2710
+    },
+    {
+      "epoch": 2.3080186677980485,
+      "grad_norm": 0.27211127699988974,
+      "learning_rate": 3.2575909979569906e-05,
+      "loss": 1.2255,
+      "step": 2720
+    },
+    {
+      "epoch": 2.316504030547306,
+      "grad_norm": 0.2741831859110416,
+      "learning_rate": 3.243458403823223e-05,
+      "loss": 1.2335,
+      "step": 2730
+    },
+    {
+      "epoch": 2.3249893932965633,
+      "grad_norm": 0.287074507507,
+      "learning_rate": 3.2292996977253075e-05,
+      "loss": 1.2555,
+      "step": 2740
+    },
+    {
+      "epoch": 2.333474756045821,
+      "grad_norm": 0.2760197579958247,
+      "learning_rate": 3.215115376949474e-05,
+      "loss": 1.2574,
+      "step": 2750
+    },
+    {
+      "epoch": 2.3419601187950785,
+      "grad_norm": 0.29917391348714156,
+      "learning_rate": 3.200905939681599e-05,
+      "loss": 1.2232,
+      "step": 2760
+    },
+    {
+      "epoch": 2.350445481544336,
+      "grad_norm": 0.2863180346672473,
+      "learning_rate": 3.1866718849897044e-05,
+      "loss": 1.2341,
+      "step": 2770
+    },
+    {
+      "epoch": 2.3589308442935937,
+      "grad_norm": 0.2760526831444543,
+      "learning_rate": 3.172413712806435e-05,
+      "loss": 1.253,
+      "step": 2780
+    },
+    {
+      "epoch": 2.3674162070428513,
+      "grad_norm": 0.29286413736773825,
+      "learning_rate": 3.158131923911498e-05,
+      "loss": 1.2617,
+      "step": 2790
+    },
+    {
+      "epoch": 2.3759015697921084,
+      "grad_norm": 0.27643034174892955,
+      "learning_rate": 3.143827019914072e-05,
+      "loss": 1.2152,
+      "step": 2800
+    },
+    {
+      "epoch": 2.384386932541366,
+      "grad_norm": 0.2939949433037669,
+      "learning_rate": 3.12949950323519e-05,
+      "loss": 1.2354,
+      "step": 2810
+    },
+    {
+      "epoch": 2.3928722952906236,
+      "grad_norm": 0.2864245267570891,
+      "learning_rate": 3.115149877090097e-05,
+      "loss": 1.2447,
+      "step": 2820
+    },
+    {
+      "epoch": 2.401357658039881,
+      "grad_norm": 0.2952829920235313,
+      "learning_rate": 3.1007786454705724e-05,
+      "loss": 1.2462,
+      "step": 2830
+    },
+    {
+      "epoch": 2.409843020789139,
+      "grad_norm": 0.3032080033620836,
+      "learning_rate": 3.0863863131272265e-05,
+      "loss": 1.2317,
+      "step": 2840
+    },
+    {
+      "epoch": 2.4183283835383964,
+      "grad_norm": 0.2678380639415362,
+      "learning_rate": 3.07197338555178e-05,
+      "loss": 1.2466,
+      "step": 2850
+    },
+    {
+      "epoch": 2.426813746287654,
+      "grad_norm": 0.3000338098809928,
+      "learning_rate": 3.0575403689593016e-05,
+      "loss": 1.2469,
+      "step": 2860
+    },
+    {
+      "epoch": 2.435299109036911,
+      "grad_norm": 0.2885428511714088,
+      "learning_rate": 3.043087770270435e-05,
+      "loss": 1.241,
+      "step": 2870
+    },
+    {
+      "epoch": 2.4437844717861688,
+      "grad_norm": 0.2902606566366597,
+      "learning_rate": 3.0286160970935906e-05,
+      "loss": 1.2498,
+      "step": 2880
+    },
+    {
+      "epoch": 2.4522698345354264,
+      "grad_norm": 0.2930924599960876,
+      "learning_rate": 3.0141258577071184e-05,
+      "loss": 1.2508,
+      "step": 2890
+    },
+    {
+      "epoch": 2.460755197284684,
+      "grad_norm": 0.28761403953538467,
+      "learning_rate": 2.9996175610414572e-05,
+      "loss": 1.2379,
+      "step": 2900
+    },
+    {
+      "epoch": 2.4692405600339415,
+      "grad_norm": 0.28888693356528744,
+      "learning_rate": 2.9850917166612586e-05,
+      "loss": 1.2383,
+      "step": 2910
+    },
+    {
+      "epoch": 2.477725922783199,
+      "grad_norm": 0.29714323219094924,
+      "learning_rate": 2.9705488347474896e-05,
+      "loss": 1.2221,
+      "step": 2920
+    },
+    {
+      "epoch": 2.4862112855324563,
+      "grad_norm": 0.3024332099011336,
+      "learning_rate": 2.9559894260795144e-05,
+      "loss": 1.2417,
+      "step": 2930
+    },
+    {
+      "epoch": 2.494696648281714,
+      "grad_norm": 0.2900123354730048,
+      "learning_rate": 2.9414140020171554e-05,
+      "loss": 1.2543,
+      "step": 2940
+    },
+    {
+      "epoch": 2.5031820110309715,
+      "grad_norm": 0.30122390943433014,
+      "learning_rate": 2.926823074482733e-05,
+      "loss": 1.2542,
+      "step": 2950
+    },
+    {
+      "epoch": 2.511667373780229,
+      "grad_norm": 0.2860208265471049,
+      "learning_rate": 2.912217155943083e-05,
+      "loss": 1.2335,
+      "step": 2960
+    },
+    {
+      "epoch": 2.5201527365294867,
+      "grad_norm": 0.28980498979259595,
+      "learning_rate": 2.897596759391561e-05,
+      "loss": 1.2458,
+      "step": 2970
+    },
+    {
+      "epoch": 2.5286380992787443,
+      "grad_norm": 0.30074882444504475,
+      "learning_rate": 2.8829623983300242e-05,
+      "loss": 1.2498,
+      "step": 2980
+    },
+    {
+      "epoch": 2.537123462028002,
+      "grad_norm": 0.2929721105596463,
+      "learning_rate": 2.868314586750794e-05,
+      "loss": 1.2686,
+      "step": 2990
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "grad_norm": 0.291755235343187,
+      "learning_rate": 2.853653839118605e-05,
+      "loss": 1.2456,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "eval_loss": 1.4051239490509033,
+      "eval_runtime": 52.7875,
+      "eval_samples_per_second": 7.218,
+      "eval_steps_per_second": 0.909,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5540941875265166,
+      "grad_norm": 0.3056527705148328,
+      "learning_rate": 2.8389806703525383e-05,
+      "loss": 1.2321,
+      "step": 3010
+    },
+    {
+      "epoch": 2.562579550275774,
+      "grad_norm": 0.29756401069688737,
+      "learning_rate": 2.8242955958079303e-05,
+      "loss": 1.2341,
+      "step": 3020
+    },
+    {
+      "epoch": 2.571064913025032,
+      "grad_norm": 0.3077048874608071,
+      "learning_rate": 2.809599131258276e-05,
+      "loss": 1.2475,
+      "step": 3030
+    },
+    {
+      "epoch": 2.5795502757742894,
+      "grad_norm": 0.31006088313098146,
+      "learning_rate": 2.7948917928771158e-05,
+      "loss": 1.2381,
+      "step": 3040
+    },
+    {
+      "epoch": 2.588035638523547,
+      "grad_norm": 0.3086227102652305,
+      "learning_rate": 2.7801740972199014e-05,
+      "loss": 1.2386,
+      "step": 3050
+    },
+    {
+      "epoch": 2.596521001272804,
+      "grad_norm": 0.2909420805400902,
+      "learning_rate": 2.7654465612058573e-05,
+      "loss": 1.2071,
+      "step": 3060
+    },
+    {
+      "epoch": 2.6050063640220618,
+      "grad_norm": 0.30310956499188235,
+      "learning_rate": 2.7507097020998246e-05,
+      "loss": 1.2206,
+      "step": 3070
+    },
+    {
+      "epoch": 2.6134917267713194,
+      "grad_norm": 0.2873915382033808,
+      "learning_rate": 2.7359640374940904e-05,
+      "loss": 1.2346,
+      "step": 3080
+    },
+    {
+      "epoch": 2.621977089520577,
+      "grad_norm": 0.29404028686651285,
+      "learning_rate": 2.7212100852902133e-05,
+      "loss": 1.2209,
+      "step": 3090
+    },
+    {
+      "epoch": 2.6304624522698346,
+      "grad_norm": 0.2967558623710032,
+      "learning_rate": 2.7064483636808313e-05,
+      "loss": 1.2471,
+      "step": 3100
+    },
+    {
+      "epoch": 2.638947815019092,
+      "grad_norm": 0.28348844201193973,
+      "learning_rate": 2.6916793911314593e-05,
+      "loss": 1.2271,
+      "step": 3110
+    },
+    {
+      "epoch": 2.6474331777683497,
+      "grad_norm": 0.30908939180701456,
+      "learning_rate": 2.6769036863622842e-05,
+      "loss": 1.2348,
+      "step": 3120
+    },
+    {
+      "epoch": 2.6559185405176073,
+      "grad_norm": 0.296064586506253,
+      "learning_rate": 2.6621217683299437e-05,
+      "loss": 1.2118,
+      "step": 3130
+    },
+    {
+      "epoch": 2.6644039032668645,
+      "grad_norm": 0.29223118946191284,
+      "learning_rate": 2.647334156209299e-05,
+      "loss": 1.2368,
+      "step": 3140
+    },
+    {
+      "epoch": 2.672889266016122,
+      "grad_norm": 0.2974562276968823,
+      "learning_rate": 2.6325413693752004e-05,
+      "loss": 1.2392,
+      "step": 3150
+    },
+    {
+      "epoch": 2.6813746287653797,
+      "grad_norm": 0.30862646184519243,
+      "learning_rate": 2.6177439273842463e-05,
+      "loss": 1.244,
+      "step": 3160
+    },
+    {
+      "epoch": 2.6898599915146373,
+      "grad_norm": 0.2958164221091078,
+      "learning_rate": 2.602942349956536e-05,
+      "loss": 1.2377,
+      "step": 3170
+    },
+    {
+      "epoch": 2.698345354263895,
+      "grad_norm": 0.2941370782364945,
+      "learning_rate": 2.5881371569574125e-05,
+      "loss": 1.2296,
+      "step": 3180
+    },
+    {
+      "epoch": 2.7068307170131525,
+      "grad_norm": 0.2949825785995608,
+      "learning_rate": 2.5733288683792084e-05,
+      "loss": 1.2292,
+      "step": 3190
+    },
+    {
+      "epoch": 2.7153160797624096,
+      "grad_norm": 0.3020705479686205,
+      "learning_rate": 2.558518004322979e-05,
+      "loss": 1.2371,
+      "step": 3200
+    },
+    {
+      "epoch": 2.7238014425116672,
+      "grad_norm": 0.3134012668403533,
+      "learning_rate": 2.5437050849802356e-05,
+      "loss": 1.2257,
+      "step": 3210
+    },
+    {
+      "epoch": 2.732286805260925,
+      "grad_norm": 0.320012175903156,
+      "learning_rate": 2.528890630614677e-05,
+      "loss": 1.215,
+      "step": 3220
+    },
+    {
+      "epoch": 2.7407721680101824,
+      "grad_norm": 0.2936053513063533,
+      "learning_rate": 2.514075161543915e-05,
+      "loss": 1.2364,
+      "step": 3230
+    },
+    {
+      "epoch": 2.74925753075944,
+      "grad_norm": 0.30515854484741317,
+      "learning_rate": 2.499259198121201e-05,
+      "loss": 1.2117,
+      "step": 3240
+    },
+    {
+      "epoch": 2.7577428935086976,
+      "grad_norm": 0.29763910785937486,
+      "learning_rate": 2.484443260717147e-05,
+      "loss": 1.2583,
+      "step": 3250
+    },
+    {
+      "epoch": 2.766228256257955,
+      "grad_norm": 0.2975895109826329,
+      "learning_rate": 2.4696278697014538e-05,
+      "loss": 1.2153,
+      "step": 3260
+    },
+    {
+      "epoch": 2.774713619007213,
+      "grad_norm": 0.2778934357454163,
+      "learning_rate": 2.4548135454246306e-05,
+      "loss": 1.2291,
+      "step": 3270
+    },
+    {
+      "epoch": 2.78319898175647,
+      "grad_norm": 0.32172302391314234,
+      "learning_rate": 2.4400008081997196e-05,
+      "loss": 1.214,
+      "step": 3280
+    },
+    {
+      "epoch": 2.7916843445057276,
+      "grad_norm": 0.3008060855751439,
+      "learning_rate": 2.425190178284024e-05,
+      "loss": 1.234,
+      "step": 3290
+    },
+    {
+      "epoch": 2.800169707254985,
+      "grad_norm": 0.3093107834664431,
+      "learning_rate": 2.4103821758608307e-05,
+      "loss": 1.2492,
+      "step": 3300
+    },
+    {
+      "epoch": 2.8086550700042427,
+      "grad_norm": 0.30554454329766617,
+      "learning_rate": 2.3955773210211465e-05,
+      "loss": 1.2401,
+      "step": 3310
+    },
+    {
+      "epoch": 2.8171404327535003,
+      "grad_norm": 0.2997812898388053,
+      "learning_rate": 2.380776133745425e-05,
+      "loss": 1.2089,
+      "step": 3320
+    },
+    {
+      "epoch": 2.8256257955027575,
+      "grad_norm": 0.2962123179493644,
+      "learning_rate": 2.3659791338853066e-05,
+      "loss": 1.2324,
+      "step": 3330
+    },
+    {
+      "epoch": 2.834111158252015,
+      "grad_norm": 0.31328021490753843,
+      "learning_rate": 2.3511868411453623e-05,
+      "loss": 1.214,
+      "step": 3340
+    },
+    {
+      "epoch": 2.8425965210012727,
+      "grad_norm": 0.30414013736166967,
+      "learning_rate": 2.3363997750648357e-05,
+      "loss": 1.2142,
+      "step": 3350
+    },
+    {
+      "epoch": 2.8510818837505303,
+      "grad_norm": 0.3079683108015467,
+      "learning_rate": 2.3216184549994006e-05,
+      "loss": 1.2137,
+      "step": 3360
+    },
+    {
+      "epoch": 2.859567246499788,
+      "grad_norm": 0.31165537422962203,
+      "learning_rate": 2.3068434001029173e-05,
+      "loss": 1.1915,
+      "step": 3370
+    },
+    {
+      "epoch": 2.8680526092490455,
+      "grad_norm": 0.3190451157121095,
+      "learning_rate": 2.2920751293091948e-05,
+      "loss": 1.2193,
+      "step": 3380
+    },
+    {
+      "epoch": 2.876537971998303,
+      "grad_norm": 0.30247955079343214,
+      "learning_rate": 2.277314161313774e-05,
+      "loss": 1.2253,
+      "step": 3390
+    },
+    {
+      "epoch": 2.8850233347475607,
+      "grad_norm": 0.2936629891547958,
+      "learning_rate": 2.262561014555703e-05,
+      "loss": 1.2136,
+      "step": 3400
+    },
+    {
+      "epoch": 2.893508697496818,
+      "grad_norm": 0.3001872886250926,
+      "learning_rate": 2.2478162071993298e-05,
+      "loss": 1.2061,
+      "step": 3410
+    },
+    {
+      "epoch": 2.9019940602460754,
+      "grad_norm": 0.3111993397471538,
+      "learning_rate": 2.233080257116103e-05,
+      "loss": 1.2193,
+      "step": 3420
+    },
+    {
+      "epoch": 2.910479422995333,
+      "grad_norm": 0.3119972602479391,
+      "learning_rate": 2.2183536818663856e-05,
+      "loss": 1.2125,
+      "step": 3430
+    },
+    {
+      "epoch": 2.9189647857445906,
+      "grad_norm": 0.30787460159489605,
+      "learning_rate": 2.2036369986812713e-05,
+      "loss": 1.215,
+      "step": 3440
+    },
+    {
+      "epoch": 2.927450148493848,
+      "grad_norm": 0.3175794611301324,
+      "learning_rate": 2.1889307244444252e-05,
+      "loss": 1.2202,
+      "step": 3450
+    },
+    {
+      "epoch": 2.9359355112431054,
+      "grad_norm": 0.2956361362699715,
+      "learning_rate": 2.1742353756739247e-05,
+      "loss": 1.2125,
+      "step": 3460
+    },
+    {
+      "epoch": 2.944420873992363,
+      "grad_norm": 0.33942331815482624,
+      "learning_rate": 2.1595514685041205e-05,
+      "loss": 1.2173,
+      "step": 3470
+    },
+    {
+      "epoch": 2.9529062367416206,
+      "grad_norm": 0.332180923100718,
+      "learning_rate": 2.144879518667507e-05,
+      "loss": 1.2266,
+      "step": 3480
+    },
+    {
+      "epoch": 2.961391599490878,
+      "grad_norm": 0.3157083099981729,
+      "learning_rate": 2.1302200414766123e-05,
+      "loss": 1.2154,
+      "step": 3490
+    },
+    {
+      "epoch": 2.9698769622401358,
+      "grad_norm": 0.3149298205272042,
+      "learning_rate": 2.1155735518058914e-05,
+      "loss": 1.2232,
+      "step": 3500
+    },
+    {
+      "epoch": 2.9783623249893934,
+      "grad_norm": 0.3132779789709915,
+      "learning_rate": 2.100940564073653e-05,
+      "loss": 1.2299,
+      "step": 3510
+    },
+    {
+      "epoch": 2.986847687738651,
+      "grad_norm": 0.29387121758746726,
+      "learning_rate": 2.086321592223984e-05,
+      "loss": 1.2219,
+      "step": 3520
+    },
+    {
+      "epoch": 2.9953330504879085,
+      "grad_norm": 0.2954282506485496,
+      "learning_rate": 2.0717171497087014e-05,
+      "loss": 1.2321,
+      "step": 3530
+    },
+    {
+      "epoch": 3.0038184132371657,
+      "grad_norm": 0.2991107469221935,
+      "learning_rate": 2.057127749469321e-05,
+      "loss": 1.2387,
+      "step": 3540
+    },
+    {
+      "epoch": 3.0123037759864233,
+      "grad_norm": 0.3135732189536929,
+      "learning_rate": 2.042553903919036e-05,
+      "loss": 1.1058,
+      "step": 3550
+    },
+    {
+      "epoch": 3.020789138735681,
+      "grad_norm": 0.3179199807851635,
+      "learning_rate": 2.0279961249247274e-05,
+      "loss": 1.0677,
+      "step": 3560
+    },
+    {
+      "epoch": 3.0292745014849385,
+      "grad_norm": 0.3217398418422315,
+      "learning_rate": 2.0134549237889765e-05,
+      "loss": 1.0978,
+      "step": 3570
+    },
+    {
+      "epoch": 3.037759864234196,
+      "grad_norm": 0.32343352041544976,
+      "learning_rate": 1.9989308112321164e-05,
+      "loss": 1.0791,
+      "step": 3580
+    },
+    {
+      "epoch": 3.0462452269834537,
+      "grad_norm": 0.32321267638520695,
+      "learning_rate": 1.9844242973742886e-05,
+      "loss": 1.0991,
+      "step": 3590
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "grad_norm": 0.321242777420917,
+      "learning_rate": 1.9699358917175297e-05,
+      "loss": 1.104,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "eval_loss": 1.3918192386627197,
+      "eval_runtime": 52.2671,
+      "eval_samples_per_second": 7.289,
+      "eval_steps_per_second": 0.918,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0632159524819684,
+      "grad_norm": 0.3313441880623986,
+      "learning_rate": 1.9554661031278712e-05,
+      "loss": 1.081,
+      "step": 3610
+    },
+    {
+      "epoch": 3.071701315231226,
+      "grad_norm": 0.3866269219140372,
+      "learning_rate": 1.9410154398174742e-05,
+      "loss": 1.0826,
+      "step": 3620
+    },
+    {
+      "epoch": 3.0801866779804836,
+      "grad_norm": 0.32719602353062216,
+      "learning_rate": 1.9265844093267728e-05,
+      "loss": 1.0934,
+      "step": 3630
+    },
+    {
+      "epoch": 3.088672040729741,
+      "grad_norm": 0.3301678171750988,
+      "learning_rate": 1.9121735185066537e-05,
+      "loss": 1.1047,
+      "step": 3640
+    },
+    {
+      "epoch": 3.097157403478999,
+      "grad_norm": 0.35644796539197005,
+      "learning_rate": 1.8977832735006522e-05,
+      "loss": 1.0994,
+      "step": 3650
+    },
+    {
+      "epoch": 3.1056427662282564,
+      "grad_norm": 0.33733028940251475,
+      "learning_rate": 1.8834141797271742e-05,
+      "loss": 1.0972,
+      "step": 3660
+    },
+    {
+      "epoch": 3.114128128977514,
+      "grad_norm": 0.32548774712269085,
+      "learning_rate": 1.8690667418617462e-05,
+      "loss": 1.1046,
+      "step": 3670
+    },
+    {
+      "epoch": 3.122613491726771,
+      "grad_norm": 0.3117109384467469,
+      "learning_rate": 1.854741463819291e-05,
+      "loss": 1.0791,
+      "step": 3680
+    },
+    {
+      "epoch": 3.1310988544760288,
+      "grad_norm": 0.3240987091142989,
+      "learning_rate": 1.8404388487364242e-05,
+      "loss": 1.0824,
+      "step": 3690
+    },
+    {
+      "epoch": 3.1395842172252864,
+      "grad_norm": 0.3346401099182515,
+      "learning_rate": 1.8261593989537895e-05,
+      "loss": 1.0753,
+      "step": 3700
+    },
+    {
+      "epoch": 3.148069579974544,
+      "grad_norm": 0.3270030472928521,
+      "learning_rate": 1.81190361599841e-05,
+      "loss": 1.0934,
+      "step": 3710
+    },
+    {
+      "epoch": 3.1565549427238015,
+      "grad_norm": 0.34129438803355183,
+      "learning_rate": 1.797672000566077e-05,
+      "loss": 1.0766,
+      "step": 3720
+    },
+    {
+      "epoch": 3.165040305473059,
+      "grad_norm": 0.37057591586243926,
+      "learning_rate": 1.783465052503762e-05,
+      "loss": 1.1049,
+      "step": 3730
+    },
+    {
+      "epoch": 3.1735256682223163,
+      "grad_norm": 0.3331237337555744,
+      "learning_rate": 1.769283270792065e-05,
+      "loss": 1.0876,
+      "step": 3740
+    },
+    {
+      "epoch": 3.182011030971574,
+      "grad_norm": 0.31904083541369294,
+      "learning_rate": 1.7551271535276792e-05,
+      "loss": 1.1206,
+      "step": 3750
+    },
+    {
+      "epoch": 3.1904963937208315,
+      "grad_norm": 0.35953020953263576,
+      "learning_rate": 1.74099719790591e-05,
+      "loss": 1.0736,
+      "step": 3760
+    },
+    {
+      "epoch": 3.198981756470089,
+      "grad_norm": 0.33595544857573634,
+      "learning_rate": 1.7268939002032035e-05,
+      "loss": 1.0969,
+      "step": 3770
+    },
+    {
+      "epoch": 3.2074671192193467,
+      "grad_norm": 0.34180259956288195,
+      "learning_rate": 1.7128177557597185e-05,
+      "loss": 1.0972,
+      "step": 3780
+    },
+    {
+      "epoch": 3.2159524819686043,
+      "grad_norm": 0.3268247890892541,
+      "learning_rate": 1.6987692589619304e-05,
+      "loss": 1.0737,
+      "step": 3790
+    },
+    {
+      "epoch": 3.224437844717862,
+      "grad_norm": 0.3207792660130559,
+      "learning_rate": 1.6847489032252627e-05,
+      "loss": 1.0797,
+      "step": 3800
+    },
+    {
+      "epoch": 3.232923207467119,
+      "grad_norm": 0.3467270107463577,
+      "learning_rate": 1.6707571809767644e-05,
+      "loss": 1.1024,
+      "step": 3810
+    },
+    {
+      "epoch": 3.2414085702163766,
+      "grad_norm": 0.35579083917156773,
+      "learning_rate": 1.656794583637807e-05,
+      "loss": 1.1026,
+      "step": 3820
+    },
+    {
+      "epoch": 3.2498939329656342,
+      "grad_norm": 0.3339161504484877,
+      "learning_rate": 1.6428616016068304e-05,
+      "loss": 1.0866,
+      "step": 3830
+    },
+    {
+      "epoch": 3.258379295714892,
+      "grad_norm": 0.34920082505086103,
+      "learning_rate": 1.628958724242117e-05,
+      "loss": 1.0789,
+      "step": 3840
+    },
+    {
+      "epoch": 3.2668646584641494,
+      "grad_norm": 0.34847114740056345,
+      "learning_rate": 1.615086439844604e-05,
+      "loss": 1.0849,
+      "step": 3850
+    },
+    {
+      "epoch": 3.275350021213407,
+      "grad_norm": 0.3164317420955498,
+      "learning_rate": 1.601245235640733e-05,
+      "loss": 1.0984,
+      "step": 3860
+    },
+    {
+      "epoch": 3.283835383962664,
+      "grad_norm": 0.32036510223283066,
+      "learning_rate": 1.5874355977653392e-05,
+      "loss": 1.1122,
+      "step": 3870
+    },
+    {
+      "epoch": 3.2923207467119218,
+      "grad_norm": 0.3618051784441363,
+      "learning_rate": 1.5736580112445738e-05,
+      "loss": 1.0942,
+      "step": 3880
+    },
+    {
+      "epoch": 3.3008061094611794,
+      "grad_norm": 0.3376776049321294,
+      "learning_rate": 1.559912959978872e-05,
+      "loss": 1.0898,
+      "step": 3890
+    },
+    {
+      "epoch": 3.309291472210437,
+      "grad_norm": 0.3402374603685398,
+      "learning_rate": 1.546200926725958e-05,
+      "loss": 1.1061,
+      "step": 3900
+    },
+    {
+      "epoch": 3.3177768349596946,
+      "grad_norm": 0.32072793842909514,
+      "learning_rate": 1.5325223930838838e-05,
+      "loss": 1.0995,
+      "step": 3910
+    },
+    {
+      "epoch": 3.326262197708952,
+      "grad_norm": 0.336633795191635,
+      "learning_rate": 1.518877839474122e-05,
+      "loss": 1.0971,
+      "step": 3920
+    },
+    {
+      "epoch": 3.3347475604582097,
+      "grad_norm": 0.36554031780441654,
+      "learning_rate": 1.5052677451246877e-05,
+      "loss": 1.0867,
+      "step": 3930
+    },
+    {
+      "epoch": 3.3432329232074673,
+      "grad_norm": 0.3569408014521597,
+      "learning_rate": 1.491692588053305e-05,
+      "loss": 1.0992,
+      "step": 3940
+    },
+    {
+      "epoch": 3.3517182859567245,
+      "grad_norm": 0.33844743386504,
+      "learning_rate": 1.4781528450506232e-05,
+      "loss": 1.103,
+      "step": 3950
+    },
+    {
+      "epoch": 3.360203648705982,
+      "grad_norm": 0.31581742738052115,
+      "learning_rate": 1.4646489916634687e-05,
+      "loss": 1.0843,
+      "step": 3960
+    },
+    {
+      "epoch": 3.3686890114552397,
+      "grad_norm": 0.3882340052077705,
+      "learning_rate": 1.4511815021781411e-05,
+      "loss": 1.1146,
+      "step": 3970
+    },
+    {
+      "epoch": 3.3771743742044973,
+      "grad_norm": 0.3309843437274728,
+      "learning_rate": 1.4377508496037567e-05,
+      "loss": 1.0751,
+      "step": 3980
+    },
+    {
+      "epoch": 3.385659736953755,
+      "grad_norm": 0.34050270685357675,
+      "learning_rate": 1.4243575056556355e-05,
+      "loss": 1.0995,
+      "step": 3990
+    },
+    {
+      "epoch": 3.3941450997030125,
+      "grad_norm": 0.3459308872000444,
+      "learning_rate": 1.4110019407387315e-05,
+      "loss": 1.0817,
+      "step": 4000
+    },
+    {
+      "epoch": 3.4026304624522696,
+      "grad_norm": 0.337212090487825,
+      "learning_rate": 1.3976846239311128e-05,
+      "loss": 1.1086,
+      "step": 4010
+    },
+    {
+      "epoch": 3.4111158252015272,
+      "grad_norm": 0.3351805063766743,
+      "learning_rate": 1.384406022967489e-05,
+      "loss": 1.0919,
+      "step": 4020
+    },
+    {
+      "epoch": 3.419601187950785,
+      "grad_norm": 0.3398996417599571,
+      "learning_rate": 1.3711666042227772e-05,
+      "loss": 1.0808,
+      "step": 4030
+    },
+    {
+      "epoch": 3.4280865507000424,
+      "grad_norm": 0.3585847037493432,
+      "learning_rate": 1.357966832695725e-05,
+      "loss": 1.1169,
+      "step": 4040
+    },
+    {
+      "epoch": 3.4365719134493,
+      "grad_norm": 0.3229408856119979,
+      "learning_rate": 1.3448071719925826e-05,
+      "loss": 1.1041,
+      "step": 4050
+    },
+    {
+      "epoch": 3.4450572761985576,
+      "grad_norm": 0.34474446572567957,
+      "learning_rate": 1.331688084310812e-05,
+      "loss": 1.0772,
+      "step": 4060
+    },
+    {
+      "epoch": 3.453542638947815,
+      "grad_norm": 0.3189660506507834,
+      "learning_rate": 1.3186100304228594e-05,
+      "loss": 1.1082,
+      "step": 4070
+    },
+    {
+      "epoch": 3.4620280016970724,
+      "grad_norm": 0.3609045764903581,
+      "learning_rate": 1.3055734696599686e-05,
+      "loss": 1.0815,
+      "step": 4080
+    },
+    {
+      "epoch": 3.47051336444633,
+      "grad_norm": 0.35237020522576973,
+      "learning_rate": 1.292578859896053e-05,
+      "loss": 1.1176,
+      "step": 4090
+    },
+    {
+      "epoch": 3.4789987271955876,
+      "grad_norm": 0.3189367280387807,
+      "learning_rate": 1.2796266575316069e-05,
+      "loss": 1.0826,
+      "step": 4100
+    },
+    {
+      "epoch": 3.487484089944845,
+      "grad_norm": 0.34730662545931273,
+      "learning_rate": 1.2667173174776823e-05,
+      "loss": 1.0908,
+      "step": 4110
+    },
+    {
+      "epoch": 3.4959694526941028,
+      "grad_norm": 0.34138804654797594,
+      "learning_rate": 1.2538512931399072e-05,
+      "loss": 1.0769,
+      "step": 4120
+    },
+    {
+      "epoch": 3.5044548154433603,
+      "grad_norm": 0.33424304735568794,
+      "learning_rate": 1.2410290364025623e-05,
+      "loss": 1.0795,
+      "step": 4130
+    },
+    {
+      "epoch": 3.5129401781926175,
+      "grad_norm": 0.3313504435273678,
+      "learning_rate": 1.2282509976127098e-05,
+      "loss": 1.0878,
+      "step": 4140
+    },
+    {
+      "epoch": 3.521425540941875,
+      "grad_norm": 0.3233457240542579,
+      "learning_rate": 1.215517625564376e-05,
+      "loss": 1.0892,
+      "step": 4150
+    },
+    {
+      "epoch": 3.5299109036911327,
+      "grad_norm": 0.36342717886744974,
+      "learning_rate": 1.2028293674827909e-05,
+      "loss": 1.1121,
+      "step": 4160
+    },
+    {
+      "epoch": 3.5383962664403903,
+      "grad_norm": 0.3384633320066937,
+      "learning_rate": 1.1901866690086747e-05,
+      "loss": 1.0873,
+      "step": 4170
+    },
+    {
+      "epoch": 3.546881629189648,
+      "grad_norm": 0.3478508212091021,
+      "learning_rate": 1.1775899741825947e-05,
+      "loss": 1.1074,
+      "step": 4180
+    },
+    {
+      "epoch": 3.5553669919389055,
+      "grad_norm": 0.33728300112135334,
+      "learning_rate": 1.1650397254293583e-05,
+      "loss": 1.0962,
+      "step": 4190
+    },
+    {
+      "epoch": 3.563852354688163,
+      "grad_norm": 0.3237234384231759,
+      "learning_rate": 1.1525363635424863e-05,
+      "loss": 1.0918,
+      "step": 4200
+    },
+    {
+      "epoch": 3.563852354688163,
+      "eval_loss": 1.3741682767868042,
+      "eval_runtime": 52.4151,
+      "eval_samples_per_second": 7.269,
+      "eval_steps_per_second": 0.916,
+      "step": 4200
+    },
+    {
+      "epoch": 3.5723377174374207,
+      "grad_norm": 0.35478171271989195,
+      "learning_rate": 1.1400803276687208e-05,
+      "loss": 1.0845,
+      "step": 4210
+    },
+    {
+      "epoch": 3.580823080186678,
+      "grad_norm": 0.37888228352789066,
+      "learning_rate": 1.1276720552926096e-05,
+      "loss": 1.0622,
+      "step": 4220
+    },
+    {
+      "epoch": 3.5893084429359354,
+      "grad_norm": 0.3279788866194937,
+      "learning_rate": 1.1153119822211338e-05,
+      "loss": 1.1021,
+      "step": 4230
+    },
+    {
+      "epoch": 3.597793805685193,
+      "grad_norm": 0.3240797704044235,
+      "learning_rate": 1.103000542568406e-05,
+      "loss": 1.0931,
+      "step": 4240
+    },
+    {
+      "epoch": 3.6062791684344506,
+      "grad_norm": 0.3326855671061254,
+      "learning_rate": 1.0907381687404206e-05,
+      "loss": 1.0856,
+      "step": 4250
+    },
+    {
+      "epoch": 3.614764531183708,
+      "grad_norm": 0.3667803858944268,
+      "learning_rate": 1.0785252914198676e-05,
+      "loss": 1.0987,
+      "step": 4260
+    },
+    {
+      "epoch": 3.6232498939329654,
+      "grad_norm": 0.36473787559668763,
+      "learning_rate": 1.0663623395510087e-05,
+      "loss": 1.0817,
+      "step": 4270
+    },
+    {
+      "epoch": 3.631735256682223,
+      "grad_norm": 0.3409471312744712,
+      "learning_rate": 1.0542497403246055e-05,
+      "loss": 1.0817,
+      "step": 4280
+    },
+    {
+      "epoch": 3.6402206194314806,
+      "grad_norm": 0.3432913496959211,
+      "learning_rate": 1.0421879191629227e-05,
+      "loss": 1.1028,
+      "step": 4290
+    },
+    {
+      "epoch": 3.648705982180738,
+      "grad_norm": 0.3432540649401779,
+      "learning_rate": 1.0301772997047809e-05,
+      "loss": 1.1015,
+      "step": 4300
+    },
+    {
+      "epoch": 3.6571913449299958,
+      "grad_norm": 0.33346746301829316,
+      "learning_rate": 1.0182183037906799e-05,
+      "loss": 1.0731,
+      "step": 4310
+    },
+    {
+      "epoch": 3.6656767076792534,
+      "grad_norm": 0.3413502676678206,
+      "learning_rate": 1.0063113514479809e-05,
+      "loss": 1.0712,
+      "step": 4320
+    },
+    {
+      "epoch": 3.674162070428511,
+      "grad_norm": 0.3285225243235751,
+      "learning_rate": 9.94456860876159e-06,
+      "loss": 1.0848,
+      "step": 4330
+    },
+    {
+      "epoch": 3.6826474331777685,
+      "grad_norm": 0.3388803496844761,
+      "learning_rate": 9.826552484321087e-06,
+      "loss": 1.0819,
+      "step": 4340
+    },
+    {
+      "epoch": 3.691132795927026,
+      "grad_norm": 0.33054778123683576,
+      "learning_rate": 9.709069286155231e-06,
+      "loss": 1.0865,
+      "step": 4350
+    },
+    {
+      "epoch": 3.6996181586762833,
+      "grad_norm": 0.33512423378595196,
+      "learning_rate": 9.592123140543388e-06,
+      "loss": 1.0799,
+      "step": 4360
+    },
+    {
+      "epoch": 3.708103521425541,
+      "grad_norm": 0.33282736103466287,
+      "learning_rate": 9.475718154902382e-06,
+      "loss": 1.089,
+      "step": 4370
+    },
+    {
+      "epoch": 3.7165888841747985,
+      "grad_norm": 0.33746488116219714,
+      "learning_rate": 9.359858417642266e-06,
+      "loss": 1.1047,
+      "step": 4380
+    },
+    {
+      "epoch": 3.725074246924056,
+      "grad_norm": 0.35251902658665213,
+      "learning_rate": 9.244547998022709e-06,
+      "loss": 1.0897,
+      "step": 4390
+    },
+    {
+      "epoch": 3.7335596096733137,
+      "grad_norm": 0.32061079556576844,
+      "learning_rate": 9.12979094601011e-06,
+      "loss": 1.0873,
+      "step": 4400
+    },
+    {
+      "epoch": 3.742044972422571,
+      "grad_norm": 0.3339779862802376,
+      "learning_rate": 9.0155912921353e-06,
+      "loss": 1.0976,
+      "step": 4410
+    },
+    {
+      "epoch": 3.7505303351718284,
+      "grad_norm": 0.33137105564248603,
+      "learning_rate": 8.901953047352032e-06,
+      "loss": 1.0815,
+      "step": 4420
+    },
+    {
+      "epoch": 3.759015697921086,
+      "grad_norm": 0.3421991932531284,
+      "learning_rate": 8.788880202896072e-06,
+      "loss": 1.0962,
+      "step": 4430
+    },
+    {
+      "epoch": 3.7675010606703436,
+      "grad_norm": 0.34520719229693314,
+      "learning_rate": 8.676376730145031e-06,
+      "loss": 1.0862,
+      "step": 4440
+    },
+    {
+      "epoch": 3.7759864234196012,
+      "grad_norm": 0.3472744469623473,
+      "learning_rate": 8.564446580478877e-06,
+      "loss": 1.0906,
+      "step": 4450
+    },
+    {
+      "epoch": 3.784471786168859,
+      "grad_norm": 0.3247372551943734,
+      "learning_rate": 8.453093685141156e-06,
+      "loss": 1.0892,
+      "step": 4460
+    },
+    {
+      "epoch": 3.7929571489181164,
+      "grad_norm": 0.32380813984188905,
+      "learning_rate": 8.342321955100935e-06,
+      "loss": 1.1084,
+      "step": 4470
+    },
+    {
+      "epoch": 3.801442511667374,
+      "grad_norm": 0.3379339847312466,
+      "learning_rate": 8.232135280915398e-06,
+      "loss": 1.0751,
+      "step": 4480
+    },
+    {
+      "epoch": 3.809927874416631,
+      "grad_norm": 0.34446270018278696,
+      "learning_rate": 8.122537532593264e-06,
+      "loss": 1.0934,
+      "step": 4490
+    },
+    {
+      "epoch": 3.8184132371658888,
+      "grad_norm": 0.34800470636637415,
+      "learning_rate": 8.013532559458761e-06,
+      "loss": 1.0896,
+      "step": 4500
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5890,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3782908946939904.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-4500/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

checkpoint-5000/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen2.5-72B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.12.0

checkpoint-5000/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-72B",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-5000/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-5000/latest

@@ -0,0 +1 @@
+global_step5000

checkpoint-5000/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-5000/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-5000/trainer_state.json

@@ -0,0 +1,3597 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.2426813746287655,
+  "eval_steps": 600,
+  "global_step": 5000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00848536274925753,
+      "grad_norm": 0.4898678891363344,
+      "learning_rate": 8.488964346349746e-07,
+      "loss": 1.8056,
+      "step": 10
+    },
+    {
+      "epoch": 0.01697072549851506,
+      "grad_norm": 0.3537473179717183,
+      "learning_rate": 1.6977928692699491e-06,
+      "loss": 1.7621,
+      "step": 20
+    },
+    {
+      "epoch": 0.025456088247772592,
+      "grad_norm": 0.28215953004159977,
+      "learning_rate": 2.546689303904924e-06,
+      "loss": 1.7571,
+      "step": 30
+    },
+    {
+      "epoch": 0.03394145099703012,
+      "grad_norm": 0.27446565146764923,
+      "learning_rate": 3.3955857385398982e-06,
+      "loss": 1.7136,
+      "step": 40
+    },
+    {
+      "epoch": 0.04242681374628765,
+      "grad_norm": 0.17051549768176558,
+      "learning_rate": 4.244482173174873e-06,
+      "loss": 1.6767,
+      "step": 50
+    },
+    {
+      "epoch": 0.050912176495545185,
+      "grad_norm": 0.17763882467320422,
+      "learning_rate": 5.093378607809848e-06,
+      "loss": 1.6371,
+      "step": 60
+    },
+    {
+      "epoch": 0.05939753924480271,
+      "grad_norm": 0.14311462596290048,
+      "learning_rate": 5.942275042444822e-06,
+      "loss": 1.6324,
+      "step": 70
+    },
+    {
+      "epoch": 0.06788290199406025,
+      "grad_norm": 0.1659540846071645,
+      "learning_rate": 6.7911714770797965e-06,
+      "loss": 1.6062,
+      "step": 80
+    },
+    {
+      "epoch": 0.07636826474331777,
+      "grad_norm": 0.20064072815620043,
+      "learning_rate": 7.640067911714771e-06,
+      "loss": 1.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0848536274925753,
+      "grad_norm": 0.2179045681711979,
+      "learning_rate": 8.488964346349745e-06,
+      "loss": 1.5898,
+      "step": 100
+    },
+    {
+      "epoch": 0.09333899024183284,
+      "grad_norm": 0.23866012053128668,
+      "learning_rate": 9.337860780984721e-06,
+      "loss": 1.5924,
+      "step": 110
+    },
+    {
+      "epoch": 0.10182435299109037,
+      "grad_norm": 0.18578051776430282,
+      "learning_rate": 1.0186757215619695e-05,
+      "loss": 1.5877,
+      "step": 120
+    },
+    {
+      "epoch": 0.1103097157403479,
+      "grad_norm": 0.2216509707409362,
+      "learning_rate": 1.103565365025467e-05,
+      "loss": 1.5947,
+      "step": 130
+    },
+    {
+      "epoch": 0.11879507848960542,
+      "grad_norm": 0.20427142255694086,
+      "learning_rate": 1.1884550084889643e-05,
+      "loss": 1.5841,
+      "step": 140
+    },
+    {
+      "epoch": 0.12728044123886295,
+      "grad_norm": 0.1765851415675038,
+      "learning_rate": 1.2733446519524619e-05,
+      "loss": 1.5878,
+      "step": 150
+    },
+    {
+      "epoch": 0.1357658039881205,
+      "grad_norm": 0.1769355117060811,
+      "learning_rate": 1.3582342954159593e-05,
+      "loss": 1.5795,
+      "step": 160
+    },
+    {
+      "epoch": 0.14425116673737803,
+      "grad_norm": 0.1617675663096666,
+      "learning_rate": 1.4431239388794569e-05,
+      "loss": 1.5549,
+      "step": 170
+    },
+    {
+      "epoch": 0.15273652948663555,
+      "grad_norm": 0.17302259072151574,
+      "learning_rate": 1.5280135823429543e-05,
+      "loss": 1.5808,
+      "step": 180
+    },
+    {
+      "epoch": 0.1612218922358931,
+      "grad_norm": 0.16876039012432806,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 1.5676,
+      "step": 190
+    },
+    {
+      "epoch": 0.1697072549851506,
+      "grad_norm": 0.19627360154037596,
+      "learning_rate": 1.697792869269949e-05,
+      "loss": 1.5598,
+      "step": 200
+    },
+    {
+      "epoch": 0.17819261773440814,
+      "grad_norm": 0.16078510362361015,
+      "learning_rate": 1.7826825127334465e-05,
+      "loss": 1.5667,
+      "step": 210
+    },
+    {
+      "epoch": 0.18667798048366568,
+      "grad_norm": 0.16044786518959703,
+      "learning_rate": 1.8675721561969442e-05,
+      "loss": 1.5815,
+      "step": 220
+    },
+    {
+      "epoch": 0.1951633432329232,
+      "grad_norm": 0.15656958873834717,
+      "learning_rate": 1.9524617996604416e-05,
+      "loss": 1.5576,
+      "step": 230
+    },
+    {
+      "epoch": 0.20364870598218074,
+      "grad_norm": 0.1687290471357602,
+      "learning_rate": 2.037351443123939e-05,
+      "loss": 1.5453,
+      "step": 240
+    },
+    {
+      "epoch": 0.21213406873143828,
+      "grad_norm": 0.1519017348276184,
+      "learning_rate": 2.1222410865874364e-05,
+      "loss": 1.5554,
+      "step": 250
+    },
+    {
+      "epoch": 0.2206194314806958,
+      "grad_norm": 0.15761892005160086,
+      "learning_rate": 2.207130730050934e-05,
+      "loss": 1.5494,
+      "step": 260
+    },
+    {
+      "epoch": 0.22910479422995333,
+      "grad_norm": 0.16857088482977495,
+      "learning_rate": 2.2920203735144312e-05,
+      "loss": 1.5794,
+      "step": 270
+    },
+    {
+      "epoch": 0.23759015697921085,
+      "grad_norm": 0.1678705209913503,
+      "learning_rate": 2.3769100169779286e-05,
+      "loss": 1.5373,
+      "step": 280
+    },
+    {
+      "epoch": 0.2460755197284684,
+      "grad_norm": 0.14812649566587394,
+      "learning_rate": 2.461799660441426e-05,
+      "loss": 1.5504,
+      "step": 290
+    },
+    {
+      "epoch": 0.2545608824777259,
+      "grad_norm": 0.17651916734325857,
+      "learning_rate": 2.5466893039049238e-05,
+      "loss": 1.5607,
+      "step": 300
+    },
+    {
+      "epoch": 0.26304624522698344,
+      "grad_norm": 0.14883055338507856,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 1.5311,
+      "step": 310
+    },
+    {
+      "epoch": 0.271531607976241,
+      "grad_norm": 0.15787522753231265,
+      "learning_rate": 2.7164685908319186e-05,
+      "loss": 1.5656,
+      "step": 320
+    },
+    {
+      "epoch": 0.2800169707254985,
+      "grad_norm": 0.1625232940237689,
+      "learning_rate": 2.801358234295416e-05,
+      "loss": 1.5686,
+      "step": 330
+    },
+    {
+      "epoch": 0.28850233347475607,
+      "grad_norm": 0.18505951289343867,
+      "learning_rate": 2.8862478777589137e-05,
+      "loss": 1.5474,
+      "step": 340
+    },
+    {
+      "epoch": 0.29698769622401355,
+      "grad_norm": 0.13785772316349984,
+      "learning_rate": 2.9711375212224108e-05,
+      "loss": 1.5696,
+      "step": 350
+    },
+    {
+      "epoch": 0.3054730589732711,
+      "grad_norm": 0.13531274658248552,
+      "learning_rate": 3.0560271646859086e-05,
+      "loss": 1.5551,
+      "step": 360
+    },
+    {
+      "epoch": 0.31395842172252864,
+      "grad_norm": 0.1366381415368909,
+      "learning_rate": 3.140916808149406e-05,
+      "loss": 1.524,
+      "step": 370
+    },
+    {
+      "epoch": 0.3224437844717862,
+      "grad_norm": 0.14587220569353926,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 1.5515,
+      "step": 380
+    },
+    {
+      "epoch": 0.3309291472210437,
+      "grad_norm": 0.13336349383744864,
+      "learning_rate": 3.310696095076401e-05,
+      "loss": 1.5457,
+      "step": 390
+    },
+    {
+      "epoch": 0.3394145099703012,
+      "grad_norm": 0.1772016947970983,
+      "learning_rate": 3.395585738539898e-05,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 0.34789987271955874,
+      "grad_norm": 0.13819420575084573,
+      "learning_rate": 3.4804753820033956e-05,
+      "loss": 1.5326,
+      "step": 410
+    },
+    {
+      "epoch": 0.3563852354688163,
+      "grad_norm": 0.12729862167862188,
+      "learning_rate": 3.565365025466893e-05,
+      "loss": 1.5387,
+      "step": 420
+    },
+    {
+      "epoch": 0.3648705982180738,
+      "grad_norm": 0.11777082851399363,
+      "learning_rate": 3.6502546689303904e-05,
+      "loss": 1.5587,
+      "step": 430
+    },
+    {
+      "epoch": 0.37335596096733137,
+      "grad_norm": 0.15372268131323022,
+      "learning_rate": 3.7351443123938885e-05,
+      "loss": 1.5362,
+      "step": 440
+    },
+    {
+      "epoch": 0.3818413237165889,
+      "grad_norm": 0.12616185572252248,
+      "learning_rate": 3.820033955857386e-05,
+      "loss": 1.5548,
+      "step": 450
+    },
+    {
+      "epoch": 0.3903266864658464,
+      "grad_norm": 0.1311200786303391,
+      "learning_rate": 3.904923599320883e-05,
+      "loss": 1.5409,
+      "step": 460
+    },
+    {
+      "epoch": 0.39881204921510394,
+      "grad_norm": 0.1707919112561785,
+      "learning_rate": 3.989813242784381e-05,
+      "loss": 1.5509,
+      "step": 470
+    },
+    {
+      "epoch": 0.4072974119643615,
+      "grad_norm": 0.14660149264284913,
+      "learning_rate": 4.074702886247878e-05,
+      "loss": 1.5433,
+      "step": 480
+    },
+    {
+      "epoch": 0.415782774713619,
+      "grad_norm": 0.12478895483834351,
+      "learning_rate": 4.1595925297113755e-05,
+      "loss": 1.5382,
+      "step": 490
+    },
+    {
+      "epoch": 0.42426813746287656,
+      "grad_norm": 0.12327957445795817,
+      "learning_rate": 4.244482173174873e-05,
+      "loss": 1.5515,
+      "step": 500
+    },
+    {
+      "epoch": 0.43275350021213405,
+      "grad_norm": 0.12922777738650987,
+      "learning_rate": 4.32937181663837e-05,
+      "loss": 1.5688,
+      "step": 510
+    },
+    {
+      "epoch": 0.4412388629613916,
+      "grad_norm": 0.12486802189783415,
+      "learning_rate": 4.414261460101868e-05,
+      "loss": 1.5452,
+      "step": 520
+    },
+    {
+      "epoch": 0.44972422571064913,
+      "grad_norm": 0.1360610874577123,
+      "learning_rate": 4.499151103565366e-05,
+      "loss": 1.5493,
+      "step": 530
+    },
+    {
+      "epoch": 0.45820958845990667,
+      "grad_norm": 0.1884897685356775,
+      "learning_rate": 4.5840407470288625e-05,
+      "loss": 1.5511,
+      "step": 540
+    },
+    {
+      "epoch": 0.4666949512091642,
+      "grad_norm": 0.12446302384809525,
+      "learning_rate": 4.6689303904923606e-05,
+      "loss": 1.5458,
+      "step": 550
+    },
+    {
+      "epoch": 0.4751803139584217,
+      "grad_norm": 0.13169591804768588,
+      "learning_rate": 4.753820033955857e-05,
+      "loss": 1.5569,
+      "step": 560
+    },
+    {
+      "epoch": 0.48366567670767924,
+      "grad_norm": 0.1343809247449631,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 1.5408,
+      "step": 570
+    },
+    {
+      "epoch": 0.4921510394569368,
+      "grad_norm": 0.14024589853602,
+      "learning_rate": 4.923599320882852e-05,
+      "loss": 1.5487,
+      "step": 580
+    },
+    {
+      "epoch": 0.5006364022061943,
+      "grad_norm": 0.16240429253875313,
+      "learning_rate": 4.999999560970061e-05,
+      "loss": 1.5488,
+      "step": 590
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "grad_norm": 0.12575424857894482,
+      "learning_rate": 4.999946877563971e-05,
+      "loss": 1.532,
+      "step": 600
+    },
+    {
+      "epoch": 0.5091217649554518,
+      "eval_loss": 1.519254446029663,
+      "eval_runtime": 53.3242,
+      "eval_samples_per_second": 7.145,
+      "eval_steps_per_second": 0.9,
+      "step": 600
+    },
+    {
+      "epoch": 0.5176071277047094,
+      "grad_norm": 0.18688482756329736,
+      "learning_rate": 4.999806390290309e-05,
+      "loss": 1.5544,
+      "step": 610
+    },
+    {
+      "epoch": 0.5260924904539669,
+      "grad_norm": 0.12425469431830571,
+      "learning_rate": 4.999578104083307e-05,
+      "loss": 1.5443,
+      "step": 620
+    },
+    {
+      "epoch": 0.5345778532032245,
+      "grad_norm": 0.1299027485420099,
+      "learning_rate": 4.999262026960902e-05,
+      "loss": 1.5569,
+      "step": 630
+    },
+    {
+      "epoch": 0.543063215952482,
+      "grad_norm": 0.11441754852508934,
+      "learning_rate": 4.998858170024449e-05,
+      "loss": 1.5316,
+      "step": 640
+    },
+    {
+      "epoch": 0.5515485787017395,
+      "grad_norm": 0.14888547248976478,
+      "learning_rate": 4.998366547458326e-05,
+      "loss": 1.5177,
+      "step": 650
+    },
+    {
+      "epoch": 0.560033941450997,
+      "grad_norm": 0.14859292774768867,
+      "learning_rate": 4.997787176529449e-05,
+      "loss": 1.5394,
+      "step": 660
+    },
+    {
+      "epoch": 0.5685193042002545,
+      "grad_norm": 0.12499154376539734,
+      "learning_rate": 4.997120077586651e-05,
+      "loss": 1.5554,
+      "step": 670
+    },
+    {
+      "epoch": 0.5770046669495121,
+      "grad_norm": 0.1218974898058821,
+      "learning_rate": 4.9963652740599774e-05,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.5854900296987696,
+      "grad_norm": 0.1273110498715124,
+      "learning_rate": 4.995522792459859e-05,
+      "loss": 1.5349,
+      "step": 690
+    },
+    {
+      "epoch": 0.5939753924480271,
+      "grad_norm": 0.12115412881719101,
+      "learning_rate": 4.994592662376183e-05,
+      "loss": 1.5419,
+      "step": 700
+    },
+    {
+      "epoch": 0.6024607551972847,
+      "grad_norm": 0.14855096330233286,
+      "learning_rate": 4.99357491647725e-05,
+      "loss": 1.513,
+      "step": 710
+    },
+    {
+      "epoch": 0.6109461179465422,
+      "grad_norm": 0.11407988659327956,
+      "learning_rate": 4.992469590508628e-05,
+      "loss": 1.5243,
+      "step": 720
+    },
+    {
+      "epoch": 0.6194314806957998,
+      "grad_norm": 0.1197712643781127,
+      "learning_rate": 4.9912767232919035e-05,
+      "loss": 1.5177,
+      "step": 730
+    },
+    {
+      "epoch": 0.6279168434450573,
+      "grad_norm": 0.12400515877262065,
+      "learning_rate": 4.9899963567233074e-05,
+      "loss": 1.5619,
+      "step": 740
+    },
+    {
+      "epoch": 0.6364022061943148,
+      "grad_norm": 0.12250385257708406,
+      "learning_rate": 4.988628535772249e-05,
+      "loss": 1.539,
+      "step": 750
+    },
+    {
+      "epoch": 0.6448875689435724,
+      "grad_norm": 0.1262441090496857,
+      "learning_rate": 4.987173308479738e-05,
+      "loss": 1.5195,
+      "step": 760
+    },
+    {
+      "epoch": 0.6533729316928298,
+      "grad_norm": 0.12459694416473029,
+      "learning_rate": 4.985630725956694e-05,
+      "loss": 1.5462,
+      "step": 770
+    },
+    {
+      "epoch": 0.6618582944420874,
+      "grad_norm": 0.12985189006106762,
+      "learning_rate": 4.9840008423821527e-05,
+      "loss": 1.5113,
+      "step": 780
+    },
+    {
+      "epoch": 0.6703436571913449,
+      "grad_norm": 0.12689306141471304,
+      "learning_rate": 4.9822837150013636e-05,
+      "loss": 1.5201,
+      "step": 790
+    },
+    {
+      "epoch": 0.6788290199406024,
+      "grad_norm": 0.15393156370587963,
+      "learning_rate": 4.980479404123778e-05,
+      "loss": 1.5121,
+      "step": 800
+    },
+    {
+      "epoch": 0.68731438268986,
+      "grad_norm": 0.13213701895207608,
+      "learning_rate": 4.978587973120931e-05,
+      "loss": 1.5307,
+      "step": 810
+    },
+    {
+      "epoch": 0.6957997454391175,
+      "grad_norm": 0.11561354931316294,
+      "learning_rate": 4.9766094884242184e-05,
+      "loss": 1.5316,
+      "step": 820
+    },
+    {
+      "epoch": 0.7042851081883751,
+      "grad_norm": 0.12414772399330044,
+      "learning_rate": 4.974544019522559e-05,
+      "loss": 1.5148,
+      "step": 830
+    },
+    {
+      "epoch": 0.7127704709376326,
+      "grad_norm": 0.1171652849153521,
+      "learning_rate": 4.972391638959959e-05,
+      "loss": 1.5096,
+      "step": 840
+    },
+    {
+      "epoch": 0.7212558336868902,
+      "grad_norm": 0.12868937349582316,
+      "learning_rate": 4.9701524223329585e-05,
+      "loss": 1.5282,
+      "step": 850
+    },
+    {
+      "epoch": 0.7297411964361477,
+      "grad_norm": 0.1200015077117309,
+      "learning_rate": 4.967826448287981e-05,
+      "loss": 1.5512,
+      "step": 860
+    },
+    {
+      "epoch": 0.7382265591854051,
+      "grad_norm": 0.12340885660045105,
+      "learning_rate": 4.96541379851857e-05,
+      "loss": 1.5394,
+      "step": 870
+    },
+    {
+      "epoch": 0.7467119219346627,
+      "grad_norm": 0.12976937691467555,
+      "learning_rate": 4.962914557762517e-05,
+      "loss": 1.51,
+      "step": 880
+    },
+    {
+      "epoch": 0.7551972846839202,
+      "grad_norm": 0.11912878476038466,
+      "learning_rate": 4.9603288137988905e-05,
+      "loss": 1.5294,
+      "step": 890
+    },
+    {
+      "epoch": 0.7636826474331778,
+      "grad_norm": 0.1299625480337927,
+      "learning_rate": 4.957656657444947e-05,
+      "loss": 1.507,
+      "step": 900
+    },
+    {
+      "epoch": 0.7721680101824353,
+      "grad_norm": 0.12380144459698468,
+      "learning_rate": 4.954898182552946e-05,
+      "loss": 1.5376,
+      "step": 910
+    },
+    {
+      "epoch": 0.7806533729316928,
+      "grad_norm": 0.13139339643682763,
+      "learning_rate": 4.9520534860068535e-05,
+      "loss": 1.5291,
+      "step": 920
+    },
+    {
+      "epoch": 0.7891387356809504,
+      "grad_norm": 0.13088956203983898,
+      "learning_rate": 4.949122667718935e-05,
+      "loss": 1.5239,
+      "step": 930
+    },
+    {
+      "epoch": 0.7976240984302079,
+      "grad_norm": 0.12586052988453703,
+      "learning_rate": 4.94610583062625e-05,
+      "loss": 1.5525,
+      "step": 940
+    },
+    {
+      "epoch": 0.8061094611794655,
+      "grad_norm": 0.12020996031652877,
+      "learning_rate": 4.943003080687035e-05,
+      "loss": 1.5525,
+      "step": 950
+    },
+    {
+      "epoch": 0.814594823928723,
+      "grad_norm": 0.12866375954060869,
+      "learning_rate": 4.9398145268769856e-05,
+      "loss": 1.5266,
+      "step": 960
+    },
+    {
+      "epoch": 0.8230801866779804,
+      "grad_norm": 0.13166136756817035,
+      "learning_rate": 4.936540281185423e-05,
+      "loss": 1.5041,
+      "step": 970
+    },
+    {
+      "epoch": 0.831565549427238,
+      "grad_norm": 0.12481946698483787,
+      "learning_rate": 4.933180458611364e-05,
+      "loss": 1.5271,
+      "step": 980
+    },
+    {
+      "epoch": 0.8400509121764955,
+      "grad_norm": 0.12264463761209114,
+      "learning_rate": 4.9297351771594844e-05,
+      "loss": 1.5354,
+      "step": 990
+    },
+    {
+      "epoch": 0.8485362749257531,
+      "grad_norm": 0.11985452856537594,
+      "learning_rate": 4.926204557835968e-05,
+      "loss": 1.5167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8570216376750106,
+      "grad_norm": 0.13125396521190327,
+      "learning_rate": 4.9225887246442634e-05,
+      "loss": 1.5282,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8655070004242681,
+      "grad_norm": 0.12730192328072554,
+      "learning_rate": 4.918887804580725e-05,
+      "loss": 1.5089,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8739923631735257,
+      "grad_norm": 0.12724644219344786,
+      "learning_rate": 4.915101927630153e-05,
+      "loss": 1.4964,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8824777259227832,
+      "grad_norm": 0.13578611501833232,
+      "learning_rate": 4.911231226761227e-05,
+      "loss": 1.5189,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8909630886720408,
+      "grad_norm": 0.13577513964986457,
+      "learning_rate": 4.90727583792184e-05,
+      "loss": 1.5149,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8994484514212983,
+      "grad_norm": 0.1269735011676505,
+      "learning_rate": 4.903235900034317e-05,
+      "loss": 1.5066,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9079338141705557,
+      "grad_norm": 0.13250058214235566,
+      "learning_rate": 4.899111554990543e-05,
+      "loss": 1.5129,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9164191769198133,
+      "grad_norm": 0.13130735246433495,
+      "learning_rate": 4.894902947646975e-05,
+      "loss": 1.5156,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9249045396690708,
+      "grad_norm": 0.1273580180253049,
+      "learning_rate": 4.890610225819553e-05,
+      "loss": 1.5324,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9333899024183284,
+      "grad_norm": 0.13155314243939242,
+      "learning_rate": 4.8862335402785136e-05,
+      "loss": 1.5106,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9418752651675859,
+      "grad_norm": 0.13564895211984299,
+      "learning_rate": 4.88177304474309e-05,
+      "loss": 1.5067,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9503606279168434,
+      "grad_norm": 0.12774735587114736,
+      "learning_rate": 4.877228895876115e-05,
+      "loss": 1.5182,
+      "step": 1120
+    },
+    {
+      "epoch": 0.958845990666101,
+      "grad_norm": 0.1307997709537685,
+      "learning_rate": 4.872601253278517e-05,
+      "loss": 1.4969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9673313534153585,
+      "grad_norm": 0.1304794845040634,
+      "learning_rate": 4.867890279483717e-05,
+      "loss": 1.5264,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9758167161646161,
+      "grad_norm": 0.13666141796489684,
+      "learning_rate": 4.8630961399519206e-05,
+      "loss": 1.5467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9843020789138736,
+      "grad_norm": 0.1370278303190263,
+      "learning_rate": 4.8582190030643e-05,
+      "loss": 1.5127,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9927874416631312,
+      "grad_norm": 0.1390936629299565,
+      "learning_rate": 4.8532590401170894e-05,
+      "loss": 1.5058,
+      "step": 1170
+    },
+    {
+      "epoch": 1.0012728044123886,
+      "grad_norm": 0.12934475548108287,
+      "learning_rate": 4.848216425315561e-05,
+      "loss": 1.5202,
+      "step": 1180
+    },
+    {
+      "epoch": 1.0097581671616462,
+      "grad_norm": 0.13898591683370803,
+      "learning_rate": 4.843091335767913e-05,
+      "loss": 1.4563,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "grad_norm": 0.17488231535826249,
+      "learning_rate": 4.837883951479043e-05,
+      "loss": 1.4402,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0182435299109036,
+      "eval_loss": 1.4955657720565796,
+      "eval_runtime": 52.424,
+      "eval_samples_per_second": 7.268,
+      "eval_steps_per_second": 0.916,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0267288926601612,
+      "grad_norm": 0.1536036344095855,
+      "learning_rate": 4.832594455344229e-05,
+      "loss": 1.4848,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0352142554094188,
+      "grad_norm": 0.15762414421336599,
+      "learning_rate": 4.827223033142706e-05,
+      "loss": 1.4567,
+      "step": 1220
+    },
+    {
+      "epoch": 1.0436996181586762,
+      "grad_norm": 0.15058229398130366,
+      "learning_rate": 4.8217698735311414e-05,
+      "loss": 1.4672,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0521849809079338,
+      "grad_norm": 0.16010992835678386,
+      "learning_rate": 4.8162351680370044e-05,
+      "loss": 1.4458,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0606703436571914,
+      "grad_norm": 0.16758816000341356,
+      "learning_rate": 4.810619111051847e-05,
+      "loss": 1.4842,
+      "step": 1250
+    },
+    {
+      "epoch": 1.069155706406449,
+      "grad_norm": 0.16559260972674986,
+      "learning_rate": 4.8049218998244696e-05,
+      "loss": 1.4556,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0776410691557063,
+      "grad_norm": 0.17237632034416966,
+      "learning_rate": 4.7991437344539966e-05,
+      "loss": 1.4813,
+      "step": 1270
+    },
+    {
+      "epoch": 1.086126431904964,
+      "grad_norm": 0.17112756741722487,
+      "learning_rate": 4.793284817882845e-05,
+      "loss": 1.4535,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0946117946542215,
+      "grad_norm": 0.16828572707718548,
+      "learning_rate": 4.787345355889604e-05,
+      "loss": 1.4344,
+      "step": 1290
+    },
+    {
+      "epoch": 1.103097157403479,
+      "grad_norm": 0.15709986047041227,
+      "learning_rate": 4.7813255570817985e-05,
+      "loss": 1.4744,
+      "step": 1300
+    },
+    {
+      "epoch": 1.1115825201527365,
+      "grad_norm": 0.16651547128146313,
+      "learning_rate": 4.775225632888568e-05,
+      "loss": 1.4561,
+      "step": 1310
+    },
+    {
+      "epoch": 1.120067882901994,
+      "grad_norm": 0.16750176017515714,
+      "learning_rate": 4.76904579755324e-05,
+      "loss": 1.4616,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1285532456512515,
+      "grad_norm": 0.1608016567554825,
+      "learning_rate": 4.7627862681258037e-05,
+      "loss": 1.4593,
+      "step": 1330
+    },
+    {
+      "epoch": 1.137038608400509,
+      "grad_norm": 0.21390766919038295,
+      "learning_rate": 4.756447264455287e-05,
+      "loss": 1.4484,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1455239711497667,
+      "grad_norm": 0.16826883293172662,
+      "learning_rate": 4.750029009182038e-05,
+      "loss": 1.4703,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1540093338990243,
+      "grad_norm": 0.17431508867079595,
+      "learning_rate": 4.7435317277299e-05,
+      "loss": 1.4701,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1624946966482816,
+      "grad_norm": 0.15973851467570443,
+      "learning_rate": 4.736955648298299e-05,
+      "loss": 1.4503,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1709800593975392,
+      "grad_norm": 0.1887713767970947,
+      "learning_rate": 4.730301001854225e-05,
+      "loss": 1.4624,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1794654221467968,
+      "grad_norm": 0.16898695344997974,
+      "learning_rate": 4.7235680221241216e-05,
+      "loss": 1.4452,
+      "step": 1390
+    },
+    {
+      "epoch": 1.1879507848960542,
+      "grad_norm": 0.20014553287073528,
+      "learning_rate": 4.716756945585681e-05,
+      "loss": 1.4717,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1964361476453118,
+      "grad_norm": 0.17137954325200072,
+      "learning_rate": 4.709868011459528e-05,
+      "loss": 1.4403,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2049215103945694,
+      "grad_norm": 0.17801721751888322,
+      "learning_rate": 4.7029014617008294e-05,
+      "loss": 1.4339,
+      "step": 1420
+    },
+    {
+      "epoch": 1.213406873143827,
+      "grad_norm": 0.17139613676642362,
+      "learning_rate": 4.695857540990789e-05,
+      "loss": 1.4573,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2218922358930844,
+      "grad_norm": 0.16971403514498054,
+      "learning_rate": 4.688736496728058e-05,
+      "loss": 1.4282,
+      "step": 1440
+    },
+    {
+      "epoch": 1.230377598642342,
+      "grad_norm": 0.17200272420880428,
+      "learning_rate": 4.681538579020038e-05,
+      "loss": 1.4434,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2388629613915996,
+      "grad_norm": 0.17208160407432616,
+      "learning_rate": 4.6742640406741106e-05,
+      "loss": 1.45,
+      "step": 1460
+    },
+    {
+      "epoch": 1.247348324140857,
+      "grad_norm": 0.1939626212901777,
+      "learning_rate": 4.666913137188743e-05,
+      "loss": 1.4608,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2558336868901145,
+      "grad_norm": 0.17291794493304186,
+      "learning_rate": 4.6594861267445236e-05,
+      "loss": 1.4671,
+      "step": 1480
+    },
+    {
+      "epoch": 1.2643190496393721,
+      "grad_norm": 0.18219792041638924,
+      "learning_rate": 4.651983270195093e-05,
+      "loss": 1.4262,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2728044123886297,
+      "grad_norm": 0.18086437830489926,
+      "learning_rate": 4.644404831057979e-05,
+      "loss": 1.4455,
+      "step": 1500
+    },
+    {
+      "epoch": 1.281289775137887,
+      "grad_norm": 0.17417619624549402,
+      "learning_rate": 4.636751075505344e-05,
+      "loss": 1.4873,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2897751378871447,
+      "grad_norm": 0.18354282411845188,
+      "learning_rate": 4.629022272354637e-05,
+      "loss": 1.4525,
+      "step": 1520
+    },
+    {
+      "epoch": 1.298260500636402,
+      "grad_norm": 0.17985617345325455,
+      "learning_rate": 4.621218693059149e-05,
+      "loss": 1.4303,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3067458633856597,
+      "grad_norm": 0.1809708317849863,
+      "learning_rate": 4.6133406116984795e-05,
+      "loss": 1.4631,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3152312261349173,
+      "grad_norm": 0.17487374671212322,
+      "learning_rate": 4.6053883049689145e-05,
+      "loss": 1.4482,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3237165888841749,
+      "grad_norm": 0.19912807671077193,
+      "learning_rate": 4.5973620521737036e-05,
+      "loss": 1.4497,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3322019516334322,
+      "grad_norm": 0.17853627546912074,
+      "learning_rate": 4.5892621352132514e-05,
+      "loss": 1.4456,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3406873143826898,
+      "grad_norm": 0.18252596927754394,
+      "learning_rate": 4.581088838575218e-05,
+      "loss": 1.4328,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3491726771319474,
+      "grad_norm": 0.17604951053556211,
+      "learning_rate": 4.572842449324525e-05,
+      "loss": 1.4442,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3576580398812048,
+      "grad_norm": 0.18358942463311748,
+      "learning_rate": 4.564523257093275e-05,
+      "loss": 1.4338,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3661434026304624,
+      "grad_norm": 0.20508703236267142,
+      "learning_rate": 4.5561315540705774e-05,
+      "loss": 1.4445,
+      "step": 1610
+    },
+    {
+      "epoch": 1.37462876537972,
+      "grad_norm": 0.18486352550747187,
+      "learning_rate": 4.547667634992288e-05,
+      "loss": 1.4261,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3831141281289776,
+      "grad_norm": 0.17492766465456316,
+      "learning_rate": 4.539131797130656e-05,
+      "loss": 1.4258,
+      "step": 1630
+    },
+    {
+      "epoch": 1.391599490878235,
+      "grad_norm": 0.19692876587833674,
+      "learning_rate": 4.530524340283881e-05,
+      "loss": 1.4349,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4000848536274926,
+      "grad_norm": 0.19155373430892478,
+      "learning_rate": 4.521845566765589e-05,
+      "loss": 1.4536,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4085702163767502,
+      "grad_norm": 0.18544325977459192,
+      "learning_rate": 4.513095781394208e-05,
+      "loss": 1.4363,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4170555791260075,
+      "grad_norm": 0.177828004720666,
+      "learning_rate": 4.504275291482267e-05,
+      "loss": 1.4595,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4255409418752651,
+      "grad_norm": 0.17855432230356816,
+      "learning_rate": 4.495384406825601e-05,
+      "loss": 1.4211,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4340263046245227,
+      "grad_norm": 0.20232492538380317,
+      "learning_rate": 4.486423439692469e-05,
+      "loss": 1.4189,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4425116673737803,
+      "grad_norm": 0.1975109303350431,
+      "learning_rate": 4.477392704812585e-05,
+      "loss": 1.4565,
+      "step": 1700
+    },
+    {
+      "epoch": 1.4509970301230377,
+      "grad_norm": 0.19619010830399825,
+      "learning_rate": 4.468292519366071e-05,
+      "loss": 1.4382,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4594823928722953,
+      "grad_norm": 0.18168826428246143,
+      "learning_rate": 4.459123202972308e-05,
+      "loss": 1.4471,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4679677556215527,
+      "grad_norm": 0.1923264062362399,
+      "learning_rate": 4.449885077678717e-05,
+      "loss": 1.4153,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4764531183708103,
+      "grad_norm": 0.1907937313040222,
+      "learning_rate": 4.440578467949445e-05,
+      "loss": 1.4432,
+      "step": 1740
+    },
+    {
+      "epoch": 1.4849384811200679,
+      "grad_norm": 0.19107457667767244,
+      "learning_rate": 4.431203700653968e-05,
+      "loss": 1.4285,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4934238438693255,
+      "grad_norm": 0.19847350429107552,
+      "learning_rate": 4.421761105055613e-05,
+      "loss": 1.4383,
+      "step": 1760
+    },
+    {
+      "epoch": 1.501909206618583,
+      "grad_norm": 0.18536475556610216,
+      "learning_rate": 4.4122510127999937e-05,
+      "loss": 1.42,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5103945693678404,
+      "grad_norm": 0.18481023473586697,
+      "learning_rate": 4.4026737579033584e-05,
+      "loss": 1.4384,
+      "step": 1780
+    },
+    {
+      "epoch": 1.518879932117098,
+      "grad_norm": 0.20863867505874642,
+      "learning_rate": 4.393029676740864e-05,
+      "loss": 1.4543,
+      "step": 1790
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "grad_norm": 0.1816036870853105,
+      "learning_rate": 4.3833191080347575e-05,
+      "loss": 1.434,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5273652948663554,
+      "eval_loss": 1.4622184038162231,
+      "eval_runtime": 52.4041,
+      "eval_samples_per_second": 7.27,
+      "eval_steps_per_second": 0.916,
+      "step": 1800
+    },
+    {
+      "epoch": 1.535850657615613,
+      "grad_norm": 0.19378252368958881,
+      "learning_rate": 4.3735423928424815e-05,
+      "loss": 1.4275,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5443360203648706,
+      "grad_norm": 0.20453331251433848,
+      "learning_rate": 4.363699874544697e-05,
+      "loss": 1.4203,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5528213831141282,
+      "grad_norm": 0.26684319417219377,
+      "learning_rate": 4.3537918988332156e-05,
+      "loss": 1.4372,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5613067458633858,
+      "grad_norm": 0.25745160303419773,
+      "learning_rate": 4.343818813698868e-05,
+      "loss": 1.4082,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5697921086126432,
+      "grad_norm": 0.19969727996700776,
+      "learning_rate": 4.3337809694192765e-05,
+      "loss": 1.4314,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5782774713619008,
+      "grad_norm": 0.20117210832277968,
+      "learning_rate": 4.3236787185465525e-05,
+      "loss": 1.4293,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5867628341111581,
+      "grad_norm": 0.20173003641028897,
+      "learning_rate": 4.313512415894913e-05,
+      "loss": 1.4406,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5952481968604157,
+      "grad_norm": 0.20304770794371527,
+      "learning_rate": 4.303282418528224e-05,
+      "loss": 1.4286,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6037335596096733,
+      "grad_norm": 0.19126658907738198,
+      "learning_rate": 4.292989085747452e-05,
+      "loss": 1.4184,
+      "step": 1890
+    },
+    {
+      "epoch": 1.612218922358931,
+      "grad_norm": 0.20069554966453027,
+      "learning_rate": 4.282632779078051e-05,
+      "loss": 1.4133,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6207042851081885,
+      "grad_norm": 0.1952881519566686,
+      "learning_rate": 4.2722138622572624e-05,
+      "loss": 1.4432,
+      "step": 1910
+    },
+    {
+      "epoch": 1.629189647857446,
+      "grad_norm": 0.19763704668680288,
+      "learning_rate": 4.261732701221339e-05,
+      "loss": 1.3921,
+      "step": 1920
+    },
+    {
+      "epoch": 1.6376750106067033,
+      "grad_norm": 0.19821464294464497,
+      "learning_rate": 4.2511896640926925e-05,
+      "loss": 1.4454,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6461603733559609,
+      "grad_norm": 0.20456545626297834,
+      "learning_rate": 4.240585121166966e-05,
+      "loss": 1.4147,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6546457361052185,
+      "grad_norm": 0.2119092529186395,
+      "learning_rate": 4.229919444900027e-05,
+      "loss": 1.3969,
+      "step": 1950
+    },
+    {
+      "epoch": 1.663131098854476,
+      "grad_norm": 0.20330157582122357,
+      "learning_rate": 4.2191930098948865e-05,
+      "loss": 1.426,
+      "step": 1960
+    },
+    {
+      "epoch": 1.6716164616037337,
+      "grad_norm": 0.21761164739298738,
+      "learning_rate": 4.2084061928885406e-05,
+      "loss": 1.4246,
+      "step": 1970
+    },
+    {
+      "epoch": 1.680101824352991,
+      "grad_norm": 0.19331588142071401,
+      "learning_rate": 4.197559372738741e-05,
+      "loss": 1.4305,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6885871871022486,
+      "grad_norm": 0.20188460724329996,
+      "learning_rate": 4.186652930410685e-05,
+      "loss": 1.4153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.697072549851506,
+      "grad_norm": 0.20988950033571588,
+      "learning_rate": 4.1756872489636425e-05,
+      "loss": 1.3894,
+      "step": 2000
+    },
+    {
+      "epoch": 1.7055579126007636,
+      "grad_norm": 0.1966475893123187,
+      "learning_rate": 4.1646627135374916e-05,
+      "loss": 1.3962,
+      "step": 2010
+    },
+    {
+      "epoch": 1.7140432753500212,
+      "grad_norm": 0.20785207367991768,
+      "learning_rate": 4.1535797113392004e-05,
+      "loss": 1.4037,
+      "step": 2020
+    },
+    {
+      "epoch": 1.7225286380992788,
+      "grad_norm": 0.2029940281663133,
+      "learning_rate": 4.1424386316292224e-05,
+      "loss": 1.4011,
+      "step": 2030
+    },
+    {
+      "epoch": 1.7310140008485364,
+      "grad_norm": 0.2247844551379277,
+      "learning_rate": 4.131239865707829e-05,
+      "loss": 1.4084,
+      "step": 2040
+    },
+    {
+      "epoch": 1.7394993635977938,
+      "grad_norm": 0.20900441746105022,
+      "learning_rate": 4.11998380690136e-05,
+      "loss": 1.4235,
+      "step": 2050
+    },
+    {
+      "epoch": 1.7479847263470514,
+      "grad_norm": 0.20362408546889926,
+      "learning_rate": 4.108670850548416e-05,
+      "loss": 1.4204,
+      "step": 2060
+    },
+    {
+      "epoch": 1.7564700890963088,
+      "grad_norm": 0.22281567946240438,
+      "learning_rate": 4.097301393985968e-05,
+      "loss": 1.4023,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7649554518455663,
+      "grad_norm": 0.20867113178797225,
+      "learning_rate": 4.085875836535404e-05,
+      "loss": 1.3895,
+      "step": 2080
+    },
+    {
+      "epoch": 1.773440814594824,
+      "grad_norm": 0.22113231886160947,
+      "learning_rate": 4.0743945794885063e-05,
+      "loss": 1.3963,
+      "step": 2090
+    },
+    {
+      "epoch": 1.7819261773440815,
+      "grad_norm": 0.22334563577844263,
+      "learning_rate": 4.062858026093351e-05,
+      "loss": 1.3988,
+      "step": 2100
+    },
+    {
+      "epoch": 1.7904115400933391,
+      "grad_norm": 0.23218581668265403,
+      "learning_rate": 4.051266581540152e-05,
+      "loss": 1.4068,
+      "step": 2110
+    },
+    {
+      "epoch": 1.7988969028425965,
+      "grad_norm": 0.20295589384571033,
+      "learning_rate": 4.0396206529470234e-05,
+      "loss": 1.3883,
+      "step": 2120
+    },
+    {
+      "epoch": 1.8073822655918539,
+      "grad_norm": 0.22861611442392848,
+      "learning_rate": 4.027920649345687e-05,
+      "loss": 1.4043,
+      "step": 2130
+    },
+    {
+      "epoch": 1.8158676283411115,
+      "grad_norm": 0.2083012771089638,
+      "learning_rate": 4.0161669816671e-05,
+      "loss": 1.398,
+      "step": 2140
+    },
+    {
+      "epoch": 1.824352991090369,
+      "grad_norm": 0.21936173231840464,
+      "learning_rate": 4.004360062727028e-05,
+      "loss": 1.4142,
+      "step": 2150
+    },
+    {
+      "epoch": 1.8328383538396267,
+      "grad_norm": 0.21383435796328337,
+      "learning_rate": 3.9925003072115406e-05,
+      "loss": 1.4138,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8413237165888843,
+      "grad_norm": 0.23301608248270392,
+      "learning_rate": 3.9805881316624506e-05,
+      "loss": 1.4195,
+      "step": 2170
+    },
+    {
+      "epoch": 1.8498090793381419,
+      "grad_norm": 0.22424766656883474,
+      "learning_rate": 3.968623954462681e-05,
+      "loss": 1.4011,
+      "step": 2180
+    },
+    {
+      "epoch": 1.8582944420873992,
+      "grad_norm": 0.21286417342881453,
+      "learning_rate": 3.9566081958215734e-05,
+      "loss": 1.409,
+      "step": 2190
+    },
+    {
+      "epoch": 1.8667798048366566,
+      "grad_norm": 0.21944800687444807,
+      "learning_rate": 3.9445412777601284e-05,
+      "loss": 1.3877,
+      "step": 2200
+    },
+    {
+      "epoch": 1.8752651675859142,
+      "grad_norm": 0.23113173625974803,
+      "learning_rate": 3.932423624096181e-05,
+      "loss": 1.4089,
+      "step": 2210
+    },
+    {
+      "epoch": 1.8837505303351718,
+      "grad_norm": 0.2081941699587778,
+      "learning_rate": 3.920255660429517e-05,
+      "loss": 1.4024,
+      "step": 2220
+    },
+    {
+      "epoch": 1.8922358930844294,
+      "grad_norm": 0.2188685806654701,
+      "learning_rate": 3.908037814126927e-05,
+      "loss": 1.3878,
+      "step": 2230
+    },
+    {
+      "epoch": 1.900721255833687,
+      "grad_norm": 0.22761843244757962,
+      "learning_rate": 3.895770514307193e-05,
+      "loss": 1.4004,
+      "step": 2240
+    },
+    {
+      "epoch": 1.9092066185829444,
+      "grad_norm": 0.23309183623120422,
+      "learning_rate": 3.883454191826017e-05,
+      "loss": 1.4188,
+      "step": 2250
+    },
+    {
+      "epoch": 1.917691981332202,
+      "grad_norm": 0.20329785843911802,
+      "learning_rate": 3.871089279260891e-05,
+      "loss": 1.3893,
+      "step": 2260
+    },
+    {
+      "epoch": 1.9261773440814594,
+      "grad_norm": 0.23470973193726366,
+      "learning_rate": 3.8586762108958995e-05,
+      "loss": 1.3974,
+      "step": 2270
+    },
+    {
+      "epoch": 1.934662706830717,
+      "grad_norm": 0.22779136837044714,
+      "learning_rate": 3.8462154227064725e-05,
+      "loss": 1.4115,
+      "step": 2280
+    },
+    {
+      "epoch": 1.9431480695799745,
+      "grad_norm": 0.22338952315651892,
+      "learning_rate": 3.833707352344068e-05,
+      "loss": 1.3873,
+      "step": 2290
+    },
+    {
+      "epoch": 1.9516334323292321,
+      "grad_norm": 0.23069304025882129,
+      "learning_rate": 3.821152439120801e-05,
+      "loss": 1.3944,
+      "step": 2300
+    },
+    {
+      "epoch": 1.9601187950784897,
+      "grad_norm": 0.23590596270163203,
+      "learning_rate": 3.808551123994018e-05,
+      "loss": 1.3857,
+      "step": 2310
+    },
+    {
+      "epoch": 1.9686041578277471,
+      "grad_norm": 0.22545661808214923,
+      "learning_rate": 3.795903849550805e-05,
+      "loss": 1.3628,
+      "step": 2320
+    },
+    {
+      "epoch": 1.9770895205770047,
+      "grad_norm": 0.2450769875954842,
+      "learning_rate": 3.7832110599924455e-05,
+      "loss": 1.4079,
+      "step": 2330
+    },
+    {
+      "epoch": 1.985574883326262,
+      "grad_norm": 0.22931499326784313,
+      "learning_rate": 3.7704732011188166e-05,
+      "loss": 1.379,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9940602460755197,
+      "grad_norm": 0.22417244507397657,
+      "learning_rate": 3.7576907203127346e-05,
+      "loss": 1.4035,
+      "step": 2350
+    },
+    {
+      "epoch": 2.0025456088247773,
+      "grad_norm": 0.24496197221575314,
+      "learning_rate": 3.7448640665242406e-05,
+      "loss": 1.442,
+      "step": 2360
+    },
+    {
+      "epoch": 2.011030971574035,
+      "grad_norm": 0.2532740296990078,
+      "learning_rate": 3.73199369025483e-05,
+      "loss": 1.2672,
+      "step": 2370
+    },
+    {
+      "epoch": 2.0195163343232925,
+      "grad_norm": 0.2890155987968593,
+      "learning_rate": 3.7190800435416355e-05,
+      "loss": 1.246,
+      "step": 2380
+    },
+    {
+      "epoch": 2.02800169707255,
+      "grad_norm": 0.2541972565696406,
+      "learning_rate": 3.706123579941545e-05,
+      "loss": 1.2603,
+      "step": 2390
+    },
+    {
+      "epoch": 2.036487059821807,
+      "grad_norm": 0.2530140862527023,
+      "learning_rate": 3.693124754515272e-05,
+      "loss": 1.2638,
+      "step": 2400
+    },
+    {
+      "epoch": 2.036487059821807,
+      "eval_loss": 1.435962438583374,
+      "eval_runtime": 52.582,
+      "eval_samples_per_second": 7.246,
+      "eval_steps_per_second": 0.913,
+      "step": 2400
+    },
+    {
+      "epoch": 2.044972422571065,
+      "grad_norm": 0.25100458343337734,
+      "learning_rate": 3.680084023811377e-05,
+      "loss": 1.2711,
+      "step": 2410
+    },
+    {
+      "epoch": 2.0534577853203224,
+      "grad_norm": 0.2695727673292618,
+      "learning_rate": 3.66700184585023e-05,
+      "loss": 1.2578,
+      "step": 2420
+    },
+    {
+      "epoch": 2.06194314806958,
+      "grad_norm": 0.2605068415443213,
+      "learning_rate": 3.6538786801079226e-05,
+      "loss": 1.2506,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0704285108188376,
+      "grad_norm": 0.27415607207865045,
+      "learning_rate": 3.64071498750013e-05,
+      "loss": 1.2852,
+      "step": 2440
+    },
+    {
+      "epoch": 2.078913873568095,
+      "grad_norm": 0.2688900338206285,
+      "learning_rate": 3.627511230365928e-05,
+      "loss": 1.2695,
+      "step": 2450
+    },
+    {
+      "epoch": 2.0873992363173524,
+      "grad_norm": 0.2750825805336503,
+      "learning_rate": 3.614267872451546e-05,
+      "loss": 1.2643,
+      "step": 2460
+    },
+    {
+      "epoch": 2.09588459906661,
+      "grad_norm": 0.2659269066581903,
+      "learning_rate": 3.600985378894086e-05,
+      "loss": 1.2868,
+      "step": 2470
+    },
+    {
+      "epoch": 2.1043699618158676,
+      "grad_norm": 0.24411151291321526,
+      "learning_rate": 3.587664216205183e-05,
+      "loss": 1.2571,
+      "step": 2480
+    },
+    {
+      "epoch": 2.112855324565125,
+      "grad_norm": 0.2574194755634052,
+      "learning_rate": 3.574304852254621e-05,
+      "loss": 1.2769,
+      "step": 2490
+    },
+    {
+      "epoch": 2.1213406873143827,
+      "grad_norm": 0.2894545074998905,
+      "learning_rate": 3.5609077562538997e-05,
+      "loss": 1.2469,
+      "step": 2500
+    },
+    {
+      "epoch": 2.1298260500636403,
+      "grad_norm": 0.2828176429904294,
+      "learning_rate": 3.547473398739754e-05,
+      "loss": 1.2527,
+      "step": 2510
+    },
+    {
+      "epoch": 2.138311412812898,
+      "grad_norm": 0.25886029771650565,
+      "learning_rate": 3.5340022515576294e-05,
+      "loss": 1.2578,
+      "step": 2520
+    },
+    {
+      "epoch": 2.146796775562155,
+      "grad_norm": 0.2783799371621383,
+      "learning_rate": 3.52049478784511e-05,
+      "loss": 1.2489,
+      "step": 2530
+    },
+    {
+      "epoch": 2.1552821383114127,
+      "grad_norm": 0.2753116113218978,
+      "learning_rate": 3.506951482015297e-05,
+      "loss": 1.275,
+      "step": 2540
+    },
+    {
+      "epoch": 2.1637675010606703,
+      "grad_norm": 0.28115792079727675,
+      "learning_rate": 3.493372809740152e-05,
+      "loss": 1.2554,
+      "step": 2550
+    },
+    {
+      "epoch": 2.172252863809928,
+      "grad_norm": 0.27954425325951715,
+      "learning_rate": 3.479759247933785e-05,
+      "loss": 1.2618,
+      "step": 2560
+    },
+    {
+      "epoch": 2.1807382265591855,
+      "grad_norm": 0.27555174232347995,
+      "learning_rate": 3.466111274735707e-05,
+      "loss": 1.2598,
+      "step": 2570
+    },
+    {
+      "epoch": 2.189223589308443,
+      "grad_norm": 0.27280827991301104,
+      "learning_rate": 3.452429369494037e-05,
+      "loss": 1.262,
+      "step": 2580
+    },
+    {
+      "epoch": 2.1977089520577007,
+      "grad_norm": 0.2749685805551003,
+      "learning_rate": 3.438714012748664e-05,
+      "loss": 1.2683,
+      "step": 2590
+    },
+    {
+      "epoch": 2.206194314806958,
+      "grad_norm": 0.2780594302788235,
+      "learning_rate": 3.424965686214371e-05,
+      "loss": 1.2462,
+      "step": 2600
+    },
+    {
+      "epoch": 2.2146796775562154,
+      "grad_norm": 0.2942257416636676,
+      "learning_rate": 3.411184872763915e-05,
+      "loss": 1.2581,
+      "step": 2610
+    },
+    {
+      "epoch": 2.223165040305473,
+      "grad_norm": 0.27000377333423803,
+      "learning_rate": 3.39737205641107e-05,
+      "loss": 1.2412,
+      "step": 2620
+    },
+    {
+      "epoch": 2.2316504030547306,
+      "grad_norm": 0.28187507810449336,
+      "learning_rate": 3.383527722293622e-05,
+      "loss": 1.2659,
+      "step": 2630
+    },
+    {
+      "epoch": 2.240135765803988,
+      "grad_norm": 0.2736213940552268,
+      "learning_rate": 3.369652356656336e-05,
+      "loss": 1.2553,
+      "step": 2640
+    },
+    {
+      "epoch": 2.248621128553246,
+      "grad_norm": 0.29698834543438446,
+      "learning_rate": 3.355746446833873e-05,
+      "loss": 1.2714,
+      "step": 2650
+    },
+    {
+      "epoch": 2.257106491302503,
+      "grad_norm": 0.2875128112484735,
+      "learning_rate": 3.3418104812336786e-05,
+      "loss": 1.2508,
+      "step": 2660
+    },
+    {
+      "epoch": 2.2655918540517606,
+      "grad_norm": 0.3016647299373059,
+      "learning_rate": 3.327844949318824e-05,
+      "loss": 1.2451,
+      "step": 2670
+    },
+    {
+      "epoch": 2.274077216801018,
+      "grad_norm": 0.27371321581702696,
+      "learning_rate": 3.3138503415908176e-05,
+      "loss": 1.2467,
+      "step": 2680
+    },
+    {
+      "epoch": 2.2825625795502757,
+      "grad_norm": 0.28374547760120017,
+      "learning_rate": 3.299827149572376e-05,
+      "loss": 1.2452,
+      "step": 2690
+    },
+    {
+      "epoch": 2.2910479422995333,
+      "grad_norm": 0.2805999278165284,
+      "learning_rate": 3.285775865790166e-05,
+      "loss": 1.2595,
+      "step": 2700
+    },
+    {
+      "epoch": 2.299533305048791,
+      "grad_norm": 0.2758019804125597,
+      "learning_rate": 3.271696983757496e-05,
+      "loss": 1.2583,
+      "step": 2710
+    },
+    {
+      "epoch": 2.3080186677980485,
+      "grad_norm": 0.27211127699988974,
+      "learning_rate": 3.2575909979569906e-05,
+      "loss": 1.2255,
+      "step": 2720
+    },
+    {
+      "epoch": 2.316504030547306,
+      "grad_norm": 0.2741831859110416,
+      "learning_rate": 3.243458403823223e-05,
+      "loss": 1.2335,
+      "step": 2730
+    },
+    {
+      "epoch": 2.3249893932965633,
+      "grad_norm": 0.287074507507,
+      "learning_rate": 3.2292996977253075e-05,
+      "loss": 1.2555,
+      "step": 2740
+    },
+    {
+      "epoch": 2.333474756045821,
+      "grad_norm": 0.2760197579958247,
+      "learning_rate": 3.215115376949474e-05,
+      "loss": 1.2574,
+      "step": 2750
+    },
+    {
+      "epoch": 2.3419601187950785,
+      "grad_norm": 0.29917391348714156,
+      "learning_rate": 3.200905939681599e-05,
+      "loss": 1.2232,
+      "step": 2760
+    },
+    {
+      "epoch": 2.350445481544336,
+      "grad_norm": 0.2863180346672473,
+      "learning_rate": 3.1866718849897044e-05,
+      "loss": 1.2341,
+      "step": 2770
+    },
+    {
+      "epoch": 2.3589308442935937,
+      "grad_norm": 0.2760526831444543,
+      "learning_rate": 3.172413712806435e-05,
+      "loss": 1.253,
+      "step": 2780
+    },
+    {
+      "epoch": 2.3674162070428513,
+      "grad_norm": 0.29286413736773825,
+      "learning_rate": 3.158131923911498e-05,
+      "loss": 1.2617,
+      "step": 2790
+    },
+    {
+      "epoch": 2.3759015697921084,
+      "grad_norm": 0.27643034174892955,
+      "learning_rate": 3.143827019914072e-05,
+      "loss": 1.2152,
+      "step": 2800
+    },
+    {
+      "epoch": 2.384386932541366,
+      "grad_norm": 0.2939949433037669,
+      "learning_rate": 3.12949950323519e-05,
+      "loss": 1.2354,
+      "step": 2810
+    },
+    {
+      "epoch": 2.3928722952906236,
+      "grad_norm": 0.2864245267570891,
+      "learning_rate": 3.115149877090097e-05,
+      "loss": 1.2447,
+      "step": 2820
+    },
+    {
+      "epoch": 2.401357658039881,
+      "grad_norm": 0.2952829920235313,
+      "learning_rate": 3.1007786454705724e-05,
+      "loss": 1.2462,
+      "step": 2830
+    },
+    {
+      "epoch": 2.409843020789139,
+      "grad_norm": 0.3032080033620836,
+      "learning_rate": 3.0863863131272265e-05,
+      "loss": 1.2317,
+      "step": 2840
+    },
+    {
+      "epoch": 2.4183283835383964,
+      "grad_norm": 0.2678380639415362,
+      "learning_rate": 3.07197338555178e-05,
+      "loss": 1.2466,
+      "step": 2850
+    },
+    {
+      "epoch": 2.426813746287654,
+      "grad_norm": 0.3000338098809928,
+      "learning_rate": 3.0575403689593016e-05,
+      "loss": 1.2469,
+      "step": 2860
+    },
+    {
+      "epoch": 2.435299109036911,
+      "grad_norm": 0.2885428511714088,
+      "learning_rate": 3.043087770270435e-05,
+      "loss": 1.241,
+      "step": 2870
+    },
+    {
+      "epoch": 2.4437844717861688,
+      "grad_norm": 0.2902606566366597,
+      "learning_rate": 3.0286160970935906e-05,
+      "loss": 1.2498,
+      "step": 2880
+    },
+    {
+      "epoch": 2.4522698345354264,
+      "grad_norm": 0.2930924599960876,
+      "learning_rate": 3.0141258577071184e-05,
+      "loss": 1.2508,
+      "step": 2890
+    },
+    {
+      "epoch": 2.460755197284684,
+      "grad_norm": 0.28761403953538467,
+      "learning_rate": 2.9996175610414572e-05,
+      "loss": 1.2379,
+      "step": 2900
+    },
+    {
+      "epoch": 2.4692405600339415,
+      "grad_norm": 0.28888693356528744,
+      "learning_rate": 2.9850917166612586e-05,
+      "loss": 1.2383,
+      "step": 2910
+    },
+    {
+      "epoch": 2.477725922783199,
+      "grad_norm": 0.29714323219094924,
+      "learning_rate": 2.9705488347474896e-05,
+      "loss": 1.2221,
+      "step": 2920
+    },
+    {
+      "epoch": 2.4862112855324563,
+      "grad_norm": 0.3024332099011336,
+      "learning_rate": 2.9559894260795144e-05,
+      "loss": 1.2417,
+      "step": 2930
+    },
+    {
+      "epoch": 2.494696648281714,
+      "grad_norm": 0.2900123354730048,
+      "learning_rate": 2.9414140020171554e-05,
+      "loss": 1.2543,
+      "step": 2940
+    },
+    {
+      "epoch": 2.5031820110309715,
+      "grad_norm": 0.30122390943433014,
+      "learning_rate": 2.926823074482733e-05,
+      "loss": 1.2542,
+      "step": 2950
+    },
+    {
+      "epoch": 2.511667373780229,
+      "grad_norm": 0.2860208265471049,
+      "learning_rate": 2.912217155943083e-05,
+      "loss": 1.2335,
+      "step": 2960
+    },
+    {
+      "epoch": 2.5201527365294867,
+      "grad_norm": 0.28980498979259595,
+      "learning_rate": 2.897596759391561e-05,
+      "loss": 1.2458,
+      "step": 2970
+    },
+    {
+      "epoch": 2.5286380992787443,
+      "grad_norm": 0.30074882444504475,
+      "learning_rate": 2.8829623983300242e-05,
+      "loss": 1.2498,
+      "step": 2980
+    },
+    {
+      "epoch": 2.537123462028002,
+      "grad_norm": 0.2929721105596463,
+      "learning_rate": 2.868314586750794e-05,
+      "loss": 1.2686,
+      "step": 2990
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "grad_norm": 0.291755235343187,
+      "learning_rate": 2.853653839118605e-05,
+      "loss": 1.2456,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5456088247772595,
+      "eval_loss": 1.4051239490509033,
+      "eval_runtime": 52.7875,
+      "eval_samples_per_second": 7.218,
+      "eval_steps_per_second": 0.909,
+      "step": 3000
+    },
+    {
+      "epoch": 2.5540941875265166,
+      "grad_norm": 0.3056527705148328,
+      "learning_rate": 2.8389806703525383e-05,
+      "loss": 1.2321,
+      "step": 3010
+    },
+    {
+      "epoch": 2.562579550275774,
+      "grad_norm": 0.29756401069688737,
+      "learning_rate": 2.8242955958079303e-05,
+      "loss": 1.2341,
+      "step": 3020
+    },
+    {
+      "epoch": 2.571064913025032,
+      "grad_norm": 0.3077048874608071,
+      "learning_rate": 2.809599131258276e-05,
+      "loss": 1.2475,
+      "step": 3030
+    },
+    {
+      "epoch": 2.5795502757742894,
+      "grad_norm": 0.31006088313098146,
+      "learning_rate": 2.7948917928771158e-05,
+      "loss": 1.2381,
+      "step": 3040
+    },
+    {
+      "epoch": 2.588035638523547,
+      "grad_norm": 0.3086227102652305,
+      "learning_rate": 2.7801740972199014e-05,
+      "loss": 1.2386,
+      "step": 3050
+    },
+    {
+      "epoch": 2.596521001272804,
+      "grad_norm": 0.2909420805400902,
+      "learning_rate": 2.7654465612058573e-05,
+      "loss": 1.2071,
+      "step": 3060
+    },
+    {
+      "epoch": 2.6050063640220618,
+      "grad_norm": 0.30310956499188235,
+      "learning_rate": 2.7507097020998246e-05,
+      "loss": 1.2206,
+      "step": 3070
+    },
+    {
+      "epoch": 2.6134917267713194,
+      "grad_norm": 0.2873915382033808,
+      "learning_rate": 2.7359640374940904e-05,
+      "loss": 1.2346,
+      "step": 3080
+    },
+    {
+      "epoch": 2.621977089520577,
+      "grad_norm": 0.29404028686651285,
+      "learning_rate": 2.7212100852902133e-05,
+      "loss": 1.2209,
+      "step": 3090
+    },
+    {
+      "epoch": 2.6304624522698346,
+      "grad_norm": 0.2967558623710032,
+      "learning_rate": 2.7064483636808313e-05,
+      "loss": 1.2471,
+      "step": 3100
+    },
+    {
+      "epoch": 2.638947815019092,
+      "grad_norm": 0.28348844201193973,
+      "learning_rate": 2.6916793911314593e-05,
+      "loss": 1.2271,
+      "step": 3110
+    },
+    {
+      "epoch": 2.6474331777683497,
+      "grad_norm": 0.30908939180701456,
+      "learning_rate": 2.6769036863622842e-05,
+      "loss": 1.2348,
+      "step": 3120
+    },
+    {
+      "epoch": 2.6559185405176073,
+      "grad_norm": 0.296064586506253,
+      "learning_rate": 2.6621217683299437e-05,
+      "loss": 1.2118,
+      "step": 3130
+    },
+    {
+      "epoch": 2.6644039032668645,
+      "grad_norm": 0.29223118946191284,
+      "learning_rate": 2.647334156209299e-05,
+      "loss": 1.2368,
+      "step": 3140
+    },
+    {
+      "epoch": 2.672889266016122,
+      "grad_norm": 0.2974562276968823,
+      "learning_rate": 2.6325413693752004e-05,
+      "loss": 1.2392,
+      "step": 3150
+    },
+    {
+      "epoch": 2.6813746287653797,
+      "grad_norm": 0.30862646184519243,
+      "learning_rate": 2.6177439273842463e-05,
+      "loss": 1.244,
+      "step": 3160
+    },
+    {
+      "epoch": 2.6898599915146373,
+      "grad_norm": 0.2958164221091078,
+      "learning_rate": 2.602942349956536e-05,
+      "loss": 1.2377,
+      "step": 3170
+    },
+    {
+      "epoch": 2.698345354263895,
+      "grad_norm": 0.2941370782364945,
+      "learning_rate": 2.5881371569574125e-05,
+      "loss": 1.2296,
+      "step": 3180
+    },
+    {
+      "epoch": 2.7068307170131525,
+      "grad_norm": 0.2949825785995608,
+      "learning_rate": 2.5733288683792084e-05,
+      "loss": 1.2292,
+      "step": 3190
+    },
+    {
+      "epoch": 2.7153160797624096,
+      "grad_norm": 0.3020705479686205,
+      "learning_rate": 2.558518004322979e-05,
+      "loss": 1.2371,
+      "step": 3200
+    },
+    {
+      "epoch": 2.7238014425116672,
+      "grad_norm": 0.3134012668403533,
+      "learning_rate": 2.5437050849802356e-05,
+      "loss": 1.2257,
+      "step": 3210
+    },
+    {
+      "epoch": 2.732286805260925,
+      "grad_norm": 0.320012175903156,
+      "learning_rate": 2.528890630614677e-05,
+      "loss": 1.215,
+      "step": 3220
+    },
+    {
+      "epoch": 2.7407721680101824,
+      "grad_norm": 0.2936053513063533,
+      "learning_rate": 2.514075161543915e-05,
+      "loss": 1.2364,
+      "step": 3230
+    },
+    {
+      "epoch": 2.74925753075944,
+      "grad_norm": 0.30515854484741317,
+      "learning_rate": 2.499259198121201e-05,
+      "loss": 1.2117,
+      "step": 3240
+    },
+    {
+      "epoch": 2.7577428935086976,
+      "grad_norm": 0.29763910785937486,
+      "learning_rate": 2.484443260717147e-05,
+      "loss": 1.2583,
+      "step": 3250
+    },
+    {
+      "epoch": 2.766228256257955,
+      "grad_norm": 0.2975895109826329,
+      "learning_rate": 2.4696278697014538e-05,
+      "loss": 1.2153,
+      "step": 3260
+    },
+    {
+      "epoch": 2.774713619007213,
+      "grad_norm": 0.2778934357454163,
+      "learning_rate": 2.4548135454246306e-05,
+      "loss": 1.2291,
+      "step": 3270
+    },
+    {
+      "epoch": 2.78319898175647,
+      "grad_norm": 0.32172302391314234,
+      "learning_rate": 2.4400008081997196e-05,
+      "loss": 1.214,
+      "step": 3280
+    },
+    {
+      "epoch": 2.7916843445057276,
+      "grad_norm": 0.3008060855751439,
+      "learning_rate": 2.425190178284024e-05,
+      "loss": 1.234,
+      "step": 3290
+    },
+    {
+      "epoch": 2.800169707254985,
+      "grad_norm": 0.3093107834664431,
+      "learning_rate": 2.4103821758608307e-05,
+      "loss": 1.2492,
+      "step": 3300
+    },
+    {
+      "epoch": 2.8086550700042427,
+      "grad_norm": 0.30554454329766617,
+      "learning_rate": 2.3955773210211465e-05,
+      "loss": 1.2401,
+      "step": 3310
+    },
+    {
+      "epoch": 2.8171404327535003,
+      "grad_norm": 0.2997812898388053,
+      "learning_rate": 2.380776133745425e-05,
+      "loss": 1.2089,
+      "step": 3320
+    },
+    {
+      "epoch": 2.8256257955027575,
+      "grad_norm": 0.2962123179493644,
+      "learning_rate": 2.3659791338853066e-05,
+      "loss": 1.2324,
+      "step": 3330
+    },
+    {
+      "epoch": 2.834111158252015,
+      "grad_norm": 0.31328021490753843,
+      "learning_rate": 2.3511868411453623e-05,
+      "loss": 1.214,
+      "step": 3340
+    },
+    {
+      "epoch": 2.8425965210012727,
+      "grad_norm": 0.30414013736166967,
+      "learning_rate": 2.3363997750648357e-05,
+      "loss": 1.2142,
+      "step": 3350
+    },
+    {
+      "epoch": 2.8510818837505303,
+      "grad_norm": 0.3079683108015467,
+      "learning_rate": 2.3216184549994006e-05,
+      "loss": 1.2137,
+      "step": 3360
+    },
+    {
+      "epoch": 2.859567246499788,
+      "grad_norm": 0.31165537422962203,
+      "learning_rate": 2.3068434001029173e-05,
+      "loss": 1.1915,
+      "step": 3370
+    },
+    {
+      "epoch": 2.8680526092490455,
+      "grad_norm": 0.3190451157121095,
+      "learning_rate": 2.2920751293091948e-05,
+      "loss": 1.2193,
+      "step": 3380
+    },
+    {
+      "epoch": 2.876537971998303,
+      "grad_norm": 0.30247955079343214,
+      "learning_rate": 2.277314161313774e-05,
+      "loss": 1.2253,
+      "step": 3390
+    },
+    {
+      "epoch": 2.8850233347475607,
+      "grad_norm": 0.2936629891547958,
+      "learning_rate": 2.262561014555703e-05,
+      "loss": 1.2136,
+      "step": 3400
+    },
+    {
+      "epoch": 2.893508697496818,
+      "grad_norm": 0.3001872886250926,
+      "learning_rate": 2.2478162071993298e-05,
+      "loss": 1.2061,
+      "step": 3410
+    },
+    {
+      "epoch": 2.9019940602460754,
+      "grad_norm": 0.3111993397471538,
+      "learning_rate": 2.233080257116103e-05,
+      "loss": 1.2193,
+      "step": 3420
+    },
+    {
+      "epoch": 2.910479422995333,
+      "grad_norm": 0.3119972602479391,
+      "learning_rate": 2.2183536818663856e-05,
+      "loss": 1.2125,
+      "step": 3430
+    },
+    {
+      "epoch": 2.9189647857445906,
+      "grad_norm": 0.30787460159489605,
+      "learning_rate": 2.2036369986812713e-05,
+      "loss": 1.215,
+      "step": 3440
+    },
+    {
+      "epoch": 2.927450148493848,
+      "grad_norm": 0.3175794611301324,
+      "learning_rate": 2.1889307244444252e-05,
+      "loss": 1.2202,
+      "step": 3450
+    },
+    {
+      "epoch": 2.9359355112431054,
+      "grad_norm": 0.2956361362699715,
+      "learning_rate": 2.1742353756739247e-05,
+      "loss": 1.2125,
+      "step": 3460
+    },
+    {
+      "epoch": 2.944420873992363,
+      "grad_norm": 0.33942331815482624,
+      "learning_rate": 2.1595514685041205e-05,
+      "loss": 1.2173,
+      "step": 3470
+    },
+    {
+      "epoch": 2.9529062367416206,
+      "grad_norm": 0.332180923100718,
+      "learning_rate": 2.144879518667507e-05,
+      "loss": 1.2266,
+      "step": 3480
+    },
+    {
+      "epoch": 2.961391599490878,
+      "grad_norm": 0.3157083099981729,
+      "learning_rate": 2.1302200414766123e-05,
+      "loss": 1.2154,
+      "step": 3490
+    },
+    {
+      "epoch": 2.9698769622401358,
+      "grad_norm": 0.3149298205272042,
+      "learning_rate": 2.1155735518058914e-05,
+      "loss": 1.2232,
+      "step": 3500
+    },
+    {
+      "epoch": 2.9783623249893934,
+      "grad_norm": 0.3132779789709915,
+      "learning_rate": 2.100940564073653e-05,
+      "loss": 1.2299,
+      "step": 3510
+    },
+    {
+      "epoch": 2.986847687738651,
+      "grad_norm": 0.29387121758746726,
+      "learning_rate": 2.086321592223984e-05,
+      "loss": 1.2219,
+      "step": 3520
+    },
+    {
+      "epoch": 2.9953330504879085,
+      "grad_norm": 0.2954282506485496,
+      "learning_rate": 2.0717171497087014e-05,
+      "loss": 1.2321,
+      "step": 3530
+    },
+    {
+      "epoch": 3.0038184132371657,
+      "grad_norm": 0.2991107469221935,
+      "learning_rate": 2.057127749469321e-05,
+      "loss": 1.2387,
+      "step": 3540
+    },
+    {
+      "epoch": 3.0123037759864233,
+      "grad_norm": 0.3135732189536929,
+      "learning_rate": 2.042553903919036e-05,
+      "loss": 1.1058,
+      "step": 3550
+    },
+    {
+      "epoch": 3.020789138735681,
+      "grad_norm": 0.3179199807851635,
+      "learning_rate": 2.0279961249247274e-05,
+      "loss": 1.0677,
+      "step": 3560
+    },
+    {
+      "epoch": 3.0292745014849385,
+      "grad_norm": 0.3217398418422315,
+      "learning_rate": 2.0134549237889765e-05,
+      "loss": 1.0978,
+      "step": 3570
+    },
+    {
+      "epoch": 3.037759864234196,
+      "grad_norm": 0.32343352041544976,
+      "learning_rate": 1.9989308112321164e-05,
+      "loss": 1.0791,
+      "step": 3580
+    },
+    {
+      "epoch": 3.0462452269834537,
+      "grad_norm": 0.32321267638520695,
+      "learning_rate": 1.9844242973742886e-05,
+      "loss": 1.0991,
+      "step": 3590
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "grad_norm": 0.321242777420917,
+      "learning_rate": 1.9699358917175297e-05,
+      "loss": 1.104,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0547305897327113,
+      "eval_loss": 1.3918192386627197,
+      "eval_runtime": 52.2671,
+      "eval_samples_per_second": 7.289,
+      "eval_steps_per_second": 0.918,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0632159524819684,
+      "grad_norm": 0.3313441880623986,
+      "learning_rate": 1.9554661031278712e-05,
+      "loss": 1.081,
+      "step": 3610
+    },
+    {
+      "epoch": 3.071701315231226,
+      "grad_norm": 0.3866269219140372,
+      "learning_rate": 1.9410154398174742e-05,
+      "loss": 1.0826,
+      "step": 3620
+    },
+    {
+      "epoch": 3.0801866779804836,
+      "grad_norm": 0.32719602353062216,
+      "learning_rate": 1.9265844093267728e-05,
+      "loss": 1.0934,
+      "step": 3630
+    },
+    {
+      "epoch": 3.088672040729741,
+      "grad_norm": 0.3301678171750988,
+      "learning_rate": 1.9121735185066537e-05,
+      "loss": 1.1047,
+      "step": 3640
+    },
+    {
+      "epoch": 3.097157403478999,
+      "grad_norm": 0.35644796539197005,
+      "learning_rate": 1.8977832735006522e-05,
+      "loss": 1.0994,
+      "step": 3650
+    },
+    {
+      "epoch": 3.1056427662282564,
+      "grad_norm": 0.33733028940251475,
+      "learning_rate": 1.8834141797271742e-05,
+      "loss": 1.0972,
+      "step": 3660
+    },
+    {
+      "epoch": 3.114128128977514,
+      "grad_norm": 0.32548774712269085,
+      "learning_rate": 1.8690667418617462e-05,
+      "loss": 1.1046,
+      "step": 3670
+    },
+    {
+      "epoch": 3.122613491726771,
+      "grad_norm": 0.3117109384467469,
+      "learning_rate": 1.854741463819291e-05,
+      "loss": 1.0791,
+      "step": 3680
+    },
+    {
+      "epoch": 3.1310988544760288,
+      "grad_norm": 0.3240987091142989,
+      "learning_rate": 1.8404388487364242e-05,
+      "loss": 1.0824,
+      "step": 3690
+    },
+    {
+      "epoch": 3.1395842172252864,
+      "grad_norm": 0.3346401099182515,
+      "learning_rate": 1.8261593989537895e-05,
+      "loss": 1.0753,
+      "step": 3700
+    },
+    {
+      "epoch": 3.148069579974544,
+      "grad_norm": 0.3270030472928521,
+      "learning_rate": 1.81190361599841e-05,
+      "loss": 1.0934,
+      "step": 3710
+    },
+    {
+      "epoch": 3.1565549427238015,
+      "grad_norm": 0.34129438803355183,
+      "learning_rate": 1.797672000566077e-05,
+      "loss": 1.0766,
+      "step": 3720
+    },
+    {
+      "epoch": 3.165040305473059,
+      "grad_norm": 0.37057591586243926,
+      "learning_rate": 1.783465052503762e-05,
+      "loss": 1.1049,
+      "step": 3730
+    },
+    {
+      "epoch": 3.1735256682223163,
+      "grad_norm": 0.3331237337555744,
+      "learning_rate": 1.769283270792065e-05,
+      "loss": 1.0876,
+      "step": 3740
+    },
+    {
+      "epoch": 3.182011030971574,
+      "grad_norm": 0.31904083541369294,
+      "learning_rate": 1.7551271535276792e-05,
+      "loss": 1.1206,
+      "step": 3750
+    },
+    {
+      "epoch": 3.1904963937208315,
+      "grad_norm": 0.35953020953263576,
+      "learning_rate": 1.74099719790591e-05,
+      "loss": 1.0736,
+      "step": 3760
+    },
+    {
+      "epoch": 3.198981756470089,
+      "grad_norm": 0.33595544857573634,
+      "learning_rate": 1.7268939002032035e-05,
+      "loss": 1.0969,
+      "step": 3770
+    },
+    {
+      "epoch": 3.2074671192193467,
+      "grad_norm": 0.34180259956288195,
+      "learning_rate": 1.7128177557597185e-05,
+      "loss": 1.0972,
+      "step": 3780
+    },
+    {
+      "epoch": 3.2159524819686043,
+      "grad_norm": 0.3268247890892541,
+      "learning_rate": 1.6987692589619304e-05,
+      "loss": 1.0737,
+      "step": 3790
+    },
+    {
+      "epoch": 3.224437844717862,
+      "grad_norm": 0.3207792660130559,
+      "learning_rate": 1.6847489032252627e-05,
+      "loss": 1.0797,
+      "step": 3800
+    },
+    {
+      "epoch": 3.232923207467119,
+      "grad_norm": 0.3467270107463577,
+      "learning_rate": 1.6707571809767644e-05,
+      "loss": 1.1024,
+      "step": 3810
+    },
+    {
+      "epoch": 3.2414085702163766,
+      "grad_norm": 0.35579083917156773,
+      "learning_rate": 1.656794583637807e-05,
+      "loss": 1.1026,
+      "step": 3820
+    },
+    {
+      "epoch": 3.2498939329656342,
+      "grad_norm": 0.3339161504484877,
+      "learning_rate": 1.6428616016068304e-05,
+      "loss": 1.0866,
+      "step": 3830
+    },
+    {
+      "epoch": 3.258379295714892,
+      "grad_norm": 0.34920082505086103,
+      "learning_rate": 1.628958724242117e-05,
+      "loss": 1.0789,
+      "step": 3840
+    },
+    {
+      "epoch": 3.2668646584641494,
+      "grad_norm": 0.34847114740056345,
+      "learning_rate": 1.615086439844604e-05,
+      "loss": 1.0849,
+      "step": 3850
+    },
+    {
+      "epoch": 3.275350021213407,
+      "grad_norm": 0.3164317420955498,
+      "learning_rate": 1.601245235640733e-05,
+      "loss": 1.0984,
+      "step": 3860
+    },
+    {
+      "epoch": 3.283835383962664,
+      "grad_norm": 0.32036510223283066,
+      "learning_rate": 1.5874355977653392e-05,
+      "loss": 1.1122,
+      "step": 3870
+    },
+    {
+      "epoch": 3.2923207467119218,
+      "grad_norm": 0.3618051784441363,
+      "learning_rate": 1.5736580112445738e-05,
+      "loss": 1.0942,
+      "step": 3880
+    },
+    {
+      "epoch": 3.3008061094611794,
+      "grad_norm": 0.3376776049321294,
+      "learning_rate": 1.559912959978872e-05,
+      "loss": 1.0898,
+      "step": 3890
+    },
+    {
+      "epoch": 3.309291472210437,
+      "grad_norm": 0.3402374603685398,
+      "learning_rate": 1.546200926725958e-05,
+      "loss": 1.1061,
+      "step": 3900
+    },
+    {
+      "epoch": 3.3177768349596946,
+      "grad_norm": 0.32072793842909514,
+      "learning_rate": 1.5325223930838838e-05,
+      "loss": 1.0995,
+      "step": 3910
+    },
+    {
+      "epoch": 3.326262197708952,
+      "grad_norm": 0.336633795191635,
+      "learning_rate": 1.518877839474122e-05,
+      "loss": 1.0971,
+      "step": 3920
+    },
+    {
+      "epoch": 3.3347475604582097,
+      "grad_norm": 0.36554031780441654,
+      "learning_rate": 1.5052677451246877e-05,
+      "loss": 1.0867,
+      "step": 3930
+    },
+    {
+      "epoch": 3.3432329232074673,
+      "grad_norm": 0.3569408014521597,
+      "learning_rate": 1.491692588053305e-05,
+      "loss": 1.0992,
+      "step": 3940
+    },
+    {
+      "epoch": 3.3517182859567245,
+      "grad_norm": 0.33844743386504,
+      "learning_rate": 1.4781528450506232e-05,
+      "loss": 1.103,
+      "step": 3950
+    },
+    {
+      "epoch": 3.360203648705982,
+      "grad_norm": 0.31581742738052115,
+      "learning_rate": 1.4646489916634687e-05,
+      "loss": 1.0843,
+      "step": 3960
+    },
+    {
+      "epoch": 3.3686890114552397,
+      "grad_norm": 0.3882340052077705,
+      "learning_rate": 1.4511815021781411e-05,
+      "loss": 1.1146,
+      "step": 3970
+    },
+    {
+      "epoch": 3.3771743742044973,
+      "grad_norm": 0.3309843437274728,
+      "learning_rate": 1.4377508496037567e-05,
+      "loss": 1.0751,
+      "step": 3980
+    },
+    {
+      "epoch": 3.385659736953755,
+      "grad_norm": 0.34050270685357675,
+      "learning_rate": 1.4243575056556355e-05,
+      "loss": 1.0995,
+      "step": 3990
+    },
+    {
+      "epoch": 3.3941450997030125,
+      "grad_norm": 0.3459308872000444,
+      "learning_rate": 1.4110019407387315e-05,
+      "loss": 1.0817,
+      "step": 4000
+    },
+    {
+      "epoch": 3.4026304624522696,
+      "grad_norm": 0.337212090487825,
+      "learning_rate": 1.3976846239311128e-05,
+      "loss": 1.1086,
+      "step": 4010
+    },
+    {
+      "epoch": 3.4111158252015272,
+      "grad_norm": 0.3351805063766743,
+      "learning_rate": 1.384406022967489e-05,
+      "loss": 1.0919,
+      "step": 4020
+    },
+    {
+      "epoch": 3.419601187950785,
+      "grad_norm": 0.3398996417599571,
+      "learning_rate": 1.3711666042227772e-05,
+      "loss": 1.0808,
+      "step": 4030
+    },
+    {
+      "epoch": 3.4280865507000424,
+      "grad_norm": 0.3585847037493432,
+      "learning_rate": 1.357966832695725e-05,
+      "loss": 1.1169,
+      "step": 4040
+    },
+    {
+      "epoch": 3.4365719134493,
+      "grad_norm": 0.3229408856119979,
+      "learning_rate": 1.3448071719925826e-05,
+      "loss": 1.1041,
+      "step": 4050
+    },
+    {
+      "epoch": 3.4450572761985576,
+      "grad_norm": 0.34474446572567957,
+      "learning_rate": 1.331688084310812e-05,
+      "loss": 1.0772,
+      "step": 4060
+    },
+    {
+      "epoch": 3.453542638947815,
+      "grad_norm": 0.3189660506507834,
+      "learning_rate": 1.3186100304228594e-05,
+      "loss": 1.1082,
+      "step": 4070
+    },
+    {
+      "epoch": 3.4620280016970724,
+      "grad_norm": 0.3609045764903581,
+      "learning_rate": 1.3055734696599686e-05,
+      "loss": 1.0815,
+      "step": 4080
+    },
+    {
+      "epoch": 3.47051336444633,
+      "grad_norm": 0.35237020522576973,
+      "learning_rate": 1.292578859896053e-05,
+      "loss": 1.1176,
+      "step": 4090
+    },
+    {
+      "epoch": 3.4789987271955876,
+      "grad_norm": 0.3189367280387807,
+      "learning_rate": 1.2796266575316069e-05,
+      "loss": 1.0826,
+      "step": 4100
+    },
+    {
+      "epoch": 3.487484089944845,
+      "grad_norm": 0.34730662545931273,
+      "learning_rate": 1.2667173174776823e-05,
+      "loss": 1.0908,
+      "step": 4110
+    },
+    {
+      "epoch": 3.4959694526941028,
+      "grad_norm": 0.34138804654797594,
+      "learning_rate": 1.2538512931399072e-05,
+      "loss": 1.0769,
+      "step": 4120
+    },
+    {
+      "epoch": 3.5044548154433603,
+      "grad_norm": 0.33424304735568794,
+      "learning_rate": 1.2410290364025623e-05,
+      "loss": 1.0795,
+      "step": 4130
+    },
+    {
+      "epoch": 3.5129401781926175,
+      "grad_norm": 0.3313504435273678,
+      "learning_rate": 1.2282509976127098e-05,
+      "loss": 1.0878,
+      "step": 4140
+    },
+    {
+      "epoch": 3.521425540941875,
+      "grad_norm": 0.3233457240542579,
+      "learning_rate": 1.215517625564376e-05,
+      "loss": 1.0892,
+      "step": 4150
+    },
+    {
+      "epoch": 3.5299109036911327,
+      "grad_norm": 0.36342717886744974,
+      "learning_rate": 1.2028293674827909e-05,
+      "loss": 1.1121,
+      "step": 4160
+    },
+    {
+      "epoch": 3.5383962664403903,
+      "grad_norm": 0.3384633320066937,
+      "learning_rate": 1.1901866690086747e-05,
+      "loss": 1.0873,
+      "step": 4170
+    },
+    {
+      "epoch": 3.546881629189648,
+      "grad_norm": 0.3478508212091021,
+      "learning_rate": 1.1775899741825947e-05,
+      "loss": 1.1074,
+      "step": 4180
+    },
+    {
+      "epoch": 3.5553669919389055,
+      "grad_norm": 0.33728300112135334,
+      "learning_rate": 1.1650397254293583e-05,
+      "loss": 1.0962,
+      "step": 4190
+    },
+    {
+      "epoch": 3.563852354688163,
+      "grad_norm": 0.3237234384231759,
+      "learning_rate": 1.1525363635424863e-05,
+      "loss": 1.0918,
+      "step": 4200
+    },
+    {
+      "epoch": 3.563852354688163,
+      "eval_loss": 1.3741682767868042,
+      "eval_runtime": 52.4151,
+      "eval_samples_per_second": 7.269,
+      "eval_steps_per_second": 0.916,
+      "step": 4200
+    },
+    {
+      "epoch": 3.5723377174374207,
+      "grad_norm": 0.35478171271989195,
+      "learning_rate": 1.1400803276687208e-05,
+      "loss": 1.0845,
+      "step": 4210
+    },
+    {
+      "epoch": 3.580823080186678,
+      "grad_norm": 0.37888228352789066,
+      "learning_rate": 1.1276720552926096e-05,
+      "loss": 1.0622,
+      "step": 4220
+    },
+    {
+      "epoch": 3.5893084429359354,
+      "grad_norm": 0.3279788866194937,
+      "learning_rate": 1.1153119822211338e-05,
+      "loss": 1.1021,
+      "step": 4230
+    },
+    {
+      "epoch": 3.597793805685193,
+      "grad_norm": 0.3240797704044235,
+      "learning_rate": 1.103000542568406e-05,
+      "loss": 1.0931,
+      "step": 4240
+    },
+    {
+      "epoch": 3.6062791684344506,
+      "grad_norm": 0.3326855671061254,
+      "learning_rate": 1.0907381687404206e-05,
+      "loss": 1.0856,
+      "step": 4250
+    },
+    {
+      "epoch": 3.614764531183708,
+      "grad_norm": 0.3667803858944268,
+      "learning_rate": 1.0785252914198676e-05,
+      "loss": 1.0987,
+      "step": 4260
+    },
+    {
+      "epoch": 3.6232498939329654,
+      "grad_norm": 0.36473787559668763,
+      "learning_rate": 1.0663623395510087e-05,
+      "loss": 1.0817,
+      "step": 4270
+    },
+    {
+      "epoch": 3.631735256682223,
+      "grad_norm": 0.3409471312744712,
+      "learning_rate": 1.0542497403246055e-05,
+      "loss": 1.0817,
+      "step": 4280
+    },
+    {
+      "epoch": 3.6402206194314806,
+      "grad_norm": 0.3432913496959211,
+      "learning_rate": 1.0421879191629227e-05,
+      "loss": 1.1028,
+      "step": 4290
+    },
+    {
+      "epoch": 3.648705982180738,
+      "grad_norm": 0.3432540649401779,
+      "learning_rate": 1.0301772997047809e-05,
+      "loss": 1.1015,
+      "step": 4300
+    },
+    {
+      "epoch": 3.6571913449299958,
+      "grad_norm": 0.33346746301829316,
+      "learning_rate": 1.0182183037906799e-05,
+      "loss": 1.0731,
+      "step": 4310
+    },
+    {
+      "epoch": 3.6656767076792534,
+      "grad_norm": 0.3413502676678206,
+      "learning_rate": 1.0063113514479809e-05,
+      "loss": 1.0712,
+      "step": 4320
+    },
+    {
+      "epoch": 3.674162070428511,
+      "grad_norm": 0.3285225243235751,
+      "learning_rate": 9.94456860876159e-06,
+      "loss": 1.0848,
+      "step": 4330
+    },
+    {
+      "epoch": 3.6826474331777685,
+      "grad_norm": 0.3388803496844761,
+      "learning_rate": 9.826552484321087e-06,
+      "loss": 1.0819,
+      "step": 4340
+    },
+    {
+      "epoch": 3.691132795927026,
+      "grad_norm": 0.33054778123683576,
+      "learning_rate": 9.709069286155231e-06,
+      "loss": 1.0865,
+      "step": 4350
+    },
+    {
+      "epoch": 3.6996181586762833,
+      "grad_norm": 0.33512423378595196,
+      "learning_rate": 9.592123140543388e-06,
+      "loss": 1.0799,
+      "step": 4360
+    },
+    {
+      "epoch": 3.708103521425541,
+      "grad_norm": 0.33282736103466287,
+      "learning_rate": 9.475718154902382e-06,
+      "loss": 1.089,
+      "step": 4370
+    },
+    {
+      "epoch": 3.7165888841747985,
+      "grad_norm": 0.33746488116219714,
+      "learning_rate": 9.359858417642266e-06,
+      "loss": 1.1047,
+      "step": 4380
+    },
+    {
+      "epoch": 3.725074246924056,
+      "grad_norm": 0.35251902658665213,
+      "learning_rate": 9.244547998022709e-06,
+      "loss": 1.0897,
+      "step": 4390
+    },
+    {
+      "epoch": 3.7335596096733137,
+      "grad_norm": 0.32061079556576844,
+      "learning_rate": 9.12979094601011e-06,
+      "loss": 1.0873,
+      "step": 4400
+    },
+    {
+      "epoch": 3.742044972422571,
+      "grad_norm": 0.3339779862802376,
+      "learning_rate": 9.0155912921353e-06,
+      "loss": 1.0976,
+      "step": 4410
+    },
+    {
+      "epoch": 3.7505303351718284,
+      "grad_norm": 0.33137105564248603,
+      "learning_rate": 8.901953047352032e-06,
+      "loss": 1.0815,
+      "step": 4420
+    },
+    {
+      "epoch": 3.759015697921086,
+      "grad_norm": 0.3421991932531284,
+      "learning_rate": 8.788880202896072e-06,
+      "loss": 1.0962,
+      "step": 4430
+    },
+    {
+      "epoch": 3.7675010606703436,
+      "grad_norm": 0.34520719229693314,
+      "learning_rate": 8.676376730145031e-06,
+      "loss": 1.0862,
+      "step": 4440
+    },
+    {
+      "epoch": 3.7759864234196012,
+      "grad_norm": 0.3472744469623473,
+      "learning_rate": 8.564446580478877e-06,
+      "loss": 1.0906,
+      "step": 4450
+    },
+    {
+      "epoch": 3.784471786168859,
+      "grad_norm": 0.3247372551943734,
+      "learning_rate": 8.453093685141156e-06,
+      "loss": 1.0892,
+      "step": 4460
+    },
+    {
+      "epoch": 3.7929571489181164,
+      "grad_norm": 0.32380813984188905,
+      "learning_rate": 8.342321955100935e-06,
+      "loss": 1.1084,
+      "step": 4470
+    },
+    {
+      "epoch": 3.801442511667374,
+      "grad_norm": 0.3379339847312466,
+      "learning_rate": 8.232135280915398e-06,
+      "loss": 1.0751,
+      "step": 4480
+    },
+    {
+      "epoch": 3.809927874416631,
+      "grad_norm": 0.34446270018278696,
+      "learning_rate": 8.122537532593264e-06,
+      "loss": 1.0934,
+      "step": 4490
+    },
+    {
+      "epoch": 3.8184132371658888,
+      "grad_norm": 0.34800470636637415,
+      "learning_rate": 8.013532559458761e-06,
+      "loss": 1.0896,
+      "step": 4500
+    },
+    {
+      "epoch": 3.8268985999151464,
+      "grad_norm": 0.34720788545028275,
+      "learning_rate": 7.90512419001656e-06,
+      "loss": 1.0755,
+      "step": 4510
+    },
+    {
+      "epoch": 3.835383962664404,
+      "grad_norm": 0.32618575043616216,
+      "learning_rate": 7.797316231817198e-06,
+      "loss": 1.0897,
+      "step": 4520
+    },
+    {
+      "epoch": 3.8438693254136616,
+      "grad_norm": 0.3301952338609731,
+      "learning_rate": 7.690112471323419e-06,
+      "loss": 1.0737,
+      "step": 4530
+    },
+    {
+      "epoch": 3.8523546881629187,
+      "grad_norm": 0.36691506455432976,
+      "learning_rate": 7.583516673777141e-06,
+      "loss": 1.0749,
+      "step": 4540
+    },
+    {
+      "epoch": 3.8608400509121763,
+      "grad_norm": 0.32038136896218583,
+      "learning_rate": 7.477532583067234e-06,
+      "loss": 1.0945,
+      "step": 4550
+    },
+    {
+      "epoch": 3.869325413661434,
+      "grad_norm": 0.3612002616743919,
+      "learning_rate": 7.3721639215980216e-06,
+      "loss": 1.0869,
+      "step": 4560
+    },
+    {
+      "epoch": 3.8778107764106915,
+      "grad_norm": 0.32001031069324903,
+      "learning_rate": 7.267414390158533e-06,
+      "loss": 1.0587,
+      "step": 4570
+    },
+    {
+      "epoch": 3.886296139159949,
+      "grad_norm": 0.3397263736848682,
+      "learning_rate": 7.163287667792557e-06,
+      "loss": 1.0815,
+      "step": 4580
+    },
+    {
+      "epoch": 3.8947815019092067,
+      "grad_norm": 0.3376479038706788,
+      "learning_rate": 7.059787411669361e-06,
+      "loss": 1.0966,
+      "step": 4590
+    },
+    {
+      "epoch": 3.9032668646584643,
+      "grad_norm": 0.3361477045154485,
+      "learning_rate": 6.9569172569553195e-06,
+      "loss": 1.095,
+      "step": 4600
+    },
+    {
+      "epoch": 3.911752227407722,
+      "grad_norm": 0.3200411096233463,
+      "learning_rate": 6.854680816686177e-06,
+      "loss": 1.0547,
+      "step": 4610
+    },
+    {
+      "epoch": 3.9202375901569795,
+      "grad_norm": 0.31629554430499285,
+      "learning_rate": 6.7530816816401745e-06,
+      "loss": 1.0722,
+      "step": 4620
+    },
+    {
+      "epoch": 3.9287229529062366,
+      "grad_norm": 0.41055357604639015,
+      "learning_rate": 6.6521234202119396e-06,
+      "loss": 1.0598,
+      "step": 4630
+    },
+    {
+      "epoch": 3.9372083156554942,
+      "grad_norm": 0.3066099811651359,
+      "learning_rate": 6.551809578287138e-06,
+      "loss": 1.0796,
+      "step": 4640
+    },
+    {
+      "epoch": 3.945693678404752,
+      "grad_norm": 0.34352050628293235,
+      "learning_rate": 6.452143679117964e-06,
+      "loss": 1.0635,
+      "step": 4650
+    },
+    {
+      "epoch": 3.9541790411540094,
+      "grad_norm": 0.33858419417369884,
+      "learning_rate": 6.353129223199353e-06,
+      "loss": 1.0939,
+      "step": 4660
+    },
+    {
+      "epoch": 3.962664403903267,
+      "grad_norm": 0.3421642104660851,
+      "learning_rate": 6.2547696881460835e-06,
+      "loss": 1.0852,
+      "step": 4670
+    },
+    {
+      "epoch": 3.971149766652524,
+      "grad_norm": 0.3146569348417928,
+      "learning_rate": 6.157068528570592e-06,
+      "loss": 1.0858,
+      "step": 4680
+    },
+    {
+      "epoch": 3.9796351294017818,
+      "grad_norm": 0.33303947321847055,
+      "learning_rate": 6.060029175961665e-06,
+      "loss": 1.0894,
+      "step": 4690
+    },
+    {
+      "epoch": 3.9881204921510394,
+      "grad_norm": 0.3246201166003683,
+      "learning_rate": 5.963655038563904e-06,
+      "loss": 1.0713,
+      "step": 4700
+    },
+    {
+      "epoch": 3.996605854900297,
+      "grad_norm": 0.33668268336604734,
+      "learning_rate": 5.867949501258047e-06,
+      "loss": 1.0714,
+      "step": 4710
+    },
+    {
+      "epoch": 4.005091217649555,
+      "grad_norm": 0.340739931299029,
+      "learning_rate": 5.7729159254420405e-06,
+      "loss": 1.1111,
+      "step": 4720
+    },
+    {
+      "epoch": 4.013576580398812,
+      "grad_norm": 0.4123963053921562,
+      "learning_rate": 5.678557648913007e-06,
+      "loss": 1.0235,
+      "step": 4730
+    },
+    {
+      "epoch": 4.02206194314807,
+      "grad_norm": 0.3503972385770886,
+      "learning_rate": 5.584877985750036e-06,
+      "loss": 1.0207,
+      "step": 4740
+    },
+    {
+      "epoch": 4.030547305897327,
+      "grad_norm": 0.36492718284781545,
+      "learning_rate": 5.491880226197707e-06,
+      "loss": 1.0065,
+      "step": 4750
+    },
+    {
+      "epoch": 4.039032668646585,
+      "grad_norm": 0.3558493920675881,
+      "learning_rate": 5.399567636550634e-06,
+      "loss": 1.012,
+      "step": 4760
+    },
+    {
+      "epoch": 4.0475180313958425,
+      "grad_norm": 0.35265155864184533,
+      "learning_rate": 5.307943459038656e-06,
+      "loss": 1.0042,
+      "step": 4770
+    },
+    {
+      "epoch": 4.0560033941451,
+      "grad_norm": 0.3628265724559059,
+      "learning_rate": 5.217010911713022e-06,
+      "loss": 1.0138,
+      "step": 4780
+    },
+    {
+      "epoch": 4.064488756894357,
+      "grad_norm": 0.366016038213362,
+      "learning_rate": 5.126773188333326e-06,
+      "loss": 1.0055,
+      "step": 4790
+    },
+    {
+      "epoch": 4.072974119643614,
+      "grad_norm": 0.3663550294337188,
+      "learning_rate": 5.037233458255375e-06,
+      "loss": 1.0143,
+      "step": 4800
+    },
+    {
+      "epoch": 4.072974119643614,
+      "eval_loss": 1.3839582204818726,
+      "eval_runtime": 52.5375,
+      "eval_samples_per_second": 7.252,
+      "eval_steps_per_second": 0.914,
+      "step": 4800
+    },
+    {
+      "epoch": 4.081459482392872,
+      "grad_norm": 0.3734159050536221,
+      "learning_rate": 4.9483948663198106e-06,
+      "loss": 1.0256,
+      "step": 4810
+    },
+    {
+      "epoch": 4.08994484514213,
+      "grad_norm": 0.3756787343497328,
+      "learning_rate": 4.860260532741739e-06,
+      "loss": 0.9939,
+      "step": 4820
+    },
+    {
+      "epoch": 4.098430207891387,
+      "grad_norm": 0.3641062929794199,
+      "learning_rate": 4.7728335530010684e-06,
+      "loss": 1.0016,
+      "step": 4830
+    },
+    {
+      "epoch": 4.106915570640645,
+      "grad_norm": 0.3720412035298521,
+      "learning_rate": 4.686116997733822e-06,
+      "loss": 1.0099,
+      "step": 4840
+    },
+    {
+      "epoch": 4.115400933389902,
+      "grad_norm": 0.3494502732717846,
+      "learning_rate": 4.600113912624305e-06,
+      "loss": 0.9903,
+      "step": 4850
+    },
+    {
+      "epoch": 4.12388629613916,
+      "grad_norm": 0.3847064809020709,
+      "learning_rate": 4.514827318298087e-06,
+      "loss": 1.0109,
+      "step": 4860
+    },
+    {
+      "epoch": 4.132371658888418,
+      "grad_norm": 0.3724702679939693,
+      "learning_rate": 4.430260210215972e-06,
+      "loss": 1.0172,
+      "step": 4870
+    },
+    {
+      "epoch": 4.140857021637675,
+      "grad_norm": 0.3854607242705416,
+      "learning_rate": 4.346415558568712e-06,
+      "loss": 1.014,
+      "step": 4880
+    },
+    {
+      "epoch": 4.149342384386933,
+      "grad_norm": 0.4119725224707018,
+      "learning_rate": 4.263296308172774e-06,
+      "loss": 1.0037,
+      "step": 4890
+    },
+    {
+      "epoch": 4.15782774713619,
+      "grad_norm": 0.38075029944179556,
+      "learning_rate": 4.180905378366845e-06,
+      "loss": 1.0004,
+      "step": 4900
+    },
+    {
+      "epoch": 4.166313109885448,
+      "grad_norm": 0.3788138960181575,
+      "learning_rate": 4.099245662909338e-06,
+      "loss": 0.9996,
+      "step": 4910
+    },
+    {
+      "epoch": 4.174798472634705,
+      "grad_norm": 0.5528304971596814,
+      "learning_rate": 4.018320029876729e-06,
+      "loss": 0.9861,
+      "step": 4920
+    },
+    {
+      "epoch": 4.183283835383962,
+      "grad_norm": 0.38085596177818104,
+      "learning_rate": 3.938131321562841e-06,
+      "loss": 1.0123,
+      "step": 4930
+    },
+    {
+      "epoch": 4.19176919813322,
+      "grad_norm": 0.4065521249170979,
+      "learning_rate": 3.858682354379012e-06,
+      "loss": 1.0124,
+      "step": 4940
+    },
+    {
+      "epoch": 4.2002545608824775,
+      "grad_norm": 0.37820891516953525,
+      "learning_rate": 3.7799759187551603e-06,
+      "loss": 0.9985,
+      "step": 4950
+    },
+    {
+      "epoch": 4.208739923631735,
+      "grad_norm": 0.38792497529901426,
+      "learning_rate": 3.7020147790418263e-06,
+      "loss": 1.0166,
+      "step": 4960
+    },
+    {
+      "epoch": 4.217225286380993,
+      "grad_norm": 0.37981175823185026,
+      "learning_rate": 3.624801673413011e-06,
+      "loss": 1.0012,
+      "step": 4970
+    },
+    {
+      "epoch": 4.22571064913025,
+      "grad_norm": 0.3752506606207182,
+      "learning_rate": 3.5483393137700717e-06,
+      "loss": 0.9926,
+      "step": 4980
+    },
+    {
+      "epoch": 4.234196011879508,
+      "grad_norm": 0.36701501672766035,
+      "learning_rate": 3.4726303856464306e-06,
+      "loss": 0.9998,
+      "step": 4990
+    },
+    {
+      "epoch": 4.2426813746287655,
+      "grad_norm": 0.37096066532390476,
+      "learning_rate": 3.3976775481132678e-06,
+      "loss": 1.004,
+      "step": 5000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5890,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4203278871035904.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}