See axolotl config

axolotl version: 0.4.1

yaml

1adapter: qlora 2base_model: unsloth/SmolLM-135M 3bf16: auto 4chat_template: llama3 5dataset_prepared_path: null 6datasets: 7- data_files: 8 - d3e5bdd73347c26c_train_data.json 9 ds_type: json 10 format: custom 11 path: /workspace/input_data/d3e5bdd73347c26c_train_data.json 12 type: 13 field_input: system_prompt 14 field_instruction: problem 15 field_output: solution 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: 128 24eval_steps: 100 25eval_table_size: null 26flash_attention: true 27fp16: null 28fsdp: null 29fsdp_config: null 30gradient_accumulation_steps: 16 31gradient_checkpointing: true 32group_by_length: true 33hub_model_id: error577/c06956a3-fe98-497b-b43e-8ab5d5a0d214 34hub_repo: null 35hub_strategy: checkpoint 36hub_token: null 37learning_rate: 0.0002 38load_in_4bit: true 39load_in_8bit: false 40local_rank: null 41logging_steps: 1 42lora_alpha: 128 43lora_dropout: 0.15 44lora_fan_in_fan_out: null 45lora_model_dir: null 46lora_r: 64 47lora_target_linear: true 48lr_scheduler: cosine 49micro_batch_size: 3 50mlflow_experiment_name: /tmp/d3e5bdd73347c26c_train_data.json 51model_type: AutoModelForCausalLM 52num_epochs: 3 53optimizer: adamw_torch_fused 54output_dir: miner_id_24 55pad_to_sequence_len: true 56restore_best_weights: true 57resume_from_checkpoint: null 58s2_attention: null 59sample_packing: false 60save_steps: 100 61sequence_len: 1024 62strict: false 63tf32: false 64tokenizer_type: AutoTokenizer 65train_on_inputs: false 66trust_remote_code: true 67val_set_size: 0.005 68wandb_entity: null 69wandb_mode: online 70wandb_name: 6f6675a6-cb87-49fd-824d-f8d0dfce8107 71wandb_project: Gradients-On-Demand 72wandb_run: your_name 73wandb_runid: 6f6675a6-cb87-49fd-824d-f8d0dfce8107 74warmup_steps: 10 75weight_decay: 0.0 76xformers_attention: null 77

c06956a3-fe98-497b-b43e-8ab5d5a0d214

• Loss: 0.9896

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0002
• train_batch_size: 3
• eval_batch_size: 3
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 48
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
• num_epochs: 3

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