See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: migtissera/Tess-v2.5-Phi-3-medium-128k-14B
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - fe297105e697bbbb_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/fe297105e697bbbb_train_data.json
  type:
    field_instruction: task
    field_output: solution
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
early_stopping_threshold: 0.001
eval_max_new_tokens: 128
eval_steps: 40
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: false
group_by_length: false
hub_model_id: mrferr3t/f42699fe-8a2a-46de-81d3-a4c4af1fea12
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0003
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 100
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
micro_batch_size: 32
mlflow_experiment_name: /tmp/fe297105e697bbbb_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 50
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
s2_attention: null
sample_packing: false
save_steps: 40
saves_per_epoch: 0
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 3fa43a59-7bfe-43c9-93ae-74585476d2fa
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 3fa43a59-7bfe-43c9-93ae-74585476d2fa
warmup_ratio: 0.05
weight_decay: 0.0
xformers_attention: null

f42699fe-8a2a-46de-81d3-a4c4af1fea12

This model is a fine-tuned version of migtissera/Tess-v2.5-Phi-3-medium-128k-14B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5931

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_bnb_8bit 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: 150
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0010	1	0.7430
No log	0.0415	40	0.6771
No log	0.0829	80	0.6422
1.3417	0.1244	120	0.6342
1.3417	0.1658	160	0.6263
1.2346	0.2073	200	0.6233
1.2346	0.2487	240	0.6192
1.2346	0.2902	280	0.6167
1.2428	0.3316	320	0.6131
1.2428	0.3731	360	0.6113
1.1918	0.4145	400	0.6059
1.1918	0.4560	440	0.6034
1.1918	0.4974	480	0.6010
1.2162	0.5389	520	0.6008
1.2162	0.5803	560	0.5946
1.1867	0.6218	600	0.5968
1.1867	0.6632	640	0.5912
1.1867	0.7047	680	0.5887
1.1763	0.7461	720	0.5855
1.1763	0.7876	760	0.5841
1.1783	0.8290	800	0.5831
1.1783	0.8705	840	0.5802
1.1783	0.9119	880	0.5773
1.1443	0.9534	920	0.5796
1.1443	0.9948	960	0.5779
1.0328	1.0363	1000	0.5931

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1