See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2base_model: NousResearch/Meta-Llama-3-8B 3bf16: auto 4chat_template: llama3 5dataset_prepared_path: null 6datasets: 7- data_files: 8 - 51051246c25f83df_train_data.json 9 ds_type: json 10 format: custom 11 path: /workspace/input_data/51051246c25f83df_train_data.json 12 type: 13 field_input: description 14 field_instruction: title 15 field_output: prompt 16 format: '{instruction} {input}' 17 no_input_format: '{instruction}' 18 system_format: '{system}' 19 system_prompt: '' 20debug: null 21deepspeed: null 22early_stopping_patience: 3 23eval_max_new_tokens: 64 24eval_steps: 25 25eval_table_size: null 26flash_attention: true 27fp16: null 28fsdp: null 29fsdp_config: null 30gradient_accumulation_steps: 8 31gradient_checkpointing: true 32group_by_length: false 33hub_model_id: tm-hf-0/a7d60c2e-816b-494e-a5cf-bf4fd157de65 34hub_repo: null 35hub_strategy: checkpoint 36hub_token: null 37learning_rate: 0.0001 38load_in_4bit: false 39load_in_8bit: false 40local_rank: null 41logging_steps: 1 42lora_alpha: 16 43lora_dropout: 0.05 44lora_fan_in_fan_out: true 45lora_model_dir: null 46lora_r: 8 47lora_target_linear: true 48lr_scheduler: cosine 49max_steps: 1000 50micro_batch_size: 2 51mlflow_experiment_name: /tmp/51051246c25f83df_train_data.json 52model_type: AutoModelForCausalLM 53num_epochs: 8 54optimizer: adamw_bnb_8bit 55output_dir: miner_id_24 56pad_to_sequence_len: true 57resume_from_checkpoint: null 58s2_attention: null 59sample_packing: false 60save_steps: 100 61sequence_len: 512 62special_tokens: 63 pad_token: <|end_of_text|> 64strict: false 65tf32: false 66tokenizer_type: AutoTokenizer 67train_on_inputs: false 68trust_remote_code: true 69val_set_size: 0.05 70wandb_entity: null 71wandb_mode: online 72wandb_name: d79944fb-5ae9-45c6-a328-4c162336fba4 73wandb_project: Gradients-On-Demand 74wandb_run: your_name 75wandb_runid: d79944fb-5ae9-45c6-a328-4c162336fba4 76warmup_steps: 100 77weight_decay: 0.01 78xformers_attention: null 79

a7d60c2e-816b-494e-a5cf-bf4fd157de65

• Loss: 0.0001

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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: 100
• training_steps: 1000

Training results

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1