End of training

README.md

@@ -47,14 +47,14 @@ flash_attention: true
 fp16: null
 fsdp: null
 fsdp_config: null
-gradient_accumulation_steps: 8
+gradient_accumulation_steps: 4
 gradient_checkpointing: true
 group_by_length: false
 hub_model_id: error577/3f83038b-0f7c-401f-b477-320813c2d642
 hub_repo: null
 hub_strategy: end
 hub_token: null
-learning_rate: 0.0001
+learning_rate: 0.0002
 load_in_4bit: true
 load_in_8bit: false
 local_rank: null
@@ -66,11 +66,11 @@ lora_model_dir: null
 lora_r: 32
 lora_target_linear: true
 lr_scheduler: cosine
-max_steps: 100
-micro_batch_size: 1
+max_steps: 500
+micro_batch_size: 2
 mlflow_experiment_name: /tmp/db79916eba4ff997_train_data.json
 model_type: AutoModelForCausalLM
-num_epochs: 1
+num_epochs: 4
 optimizer: adamw_bnb_8bit
 output_dir: miner_id_24
 pad_to_sequence_len: true
@@ -103,7 +103,7 @@ xformers_attention: null
 
 This model is a fine-tuned version of [unsloth/SmolLM-1.7B](https://huggingface.co/unsloth/SmolLM-1.7B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.0123
+- Loss: 1.6602
 
 ## Model description
 
@@ -122,22 +122,26 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 0.0001
-- train_batch_size: 1
-- eval_batch_size: 1
+- learning_rate: 0.0002
+- train_batch_size: 2
+- eval_batch_size: 2
 - seed: 42
-- gradient_accumulation_steps: 8
+- gradient_accumulation_steps: 4
 - total_train_batch_size: 8
 - optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_steps: 10
-- training_steps: 100
+- training_steps: 500
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 2.023         | 0.0207 | 100  | 2.0123          |
+| 2.6068        | 0.0002 | 1    | 2.9328          |
+| 1.4663        | 0.0259 | 125  | 1.8055          |
+| 1.7349        | 0.0518 | 250  | 1.7140          |
+| 1.6004        | 0.0776 | 375  | 1.6695          |
+| 1.3134        | 0.1035 | 500  | 1.6602          |
 
 
 ### Framework versions

adapter_config.json

@@ -20,12 +20,12 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "q_proj",
     "down_proj",
     "k_proj",
-    "gate_proj",
-    "up_proj",
     "o_proj",
-    "q_proj",
     "v_proj"
   ],
   "task_type": "CAUSAL_LM",

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed7b95649d746d4bdd6dd402a67c03d4248a82c23144097e7ee6569920fb32c0
+oid sha256:2184d3168dce96eb5894e14ae2517797d5ef3243226c3723e42f9c1bacb2aa32
 size 144824970

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48ca79489b3bf8495c0ff1ebcef649a216c871fb625c7fff3b24e5c719fba746
+oid sha256:91b657d35e3e230451966ea90ae95be9738f2090b7ea70e812e0c065e9c82986
 size 144748392

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dcca1f706771354531ef976c33d43361a116644dfb568c00352291242330cb4d
+oid sha256:923e314eb82c2038d10a0de5b06d6fa3b32d9426f9ffb404175c90c23b93899d
 size 6776