See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2base_model: unsloth/SmolLM2-135M 3bf16: auto 4chat_template: llama3 5dataset_prepared_path: null 6datasets: 7- data_files: 8 - a0682f06fa4fb877_train_data.json 9 ds_type: json 10 format: custom 11 path: /workspace/input_data/a0682f06fa4fb877_train_data.json 12 type: 13 field_instruction: instruction 14 field_output: positive_sample 15 format: '{instruction}' 16 no_input_format: '{instruction}' 17 system_format: '{system}' 18 system_prompt: '' 19debug: null 20deepspeed: null 21early_stopping_patience: null 22eval_max_new_tokens: 128 23eval_table_size: null 24evals_per_epoch: 1 25flash_attention: true 26fp16: null 27fsdp: null 28fsdp_config: null 29gradient_accumulation_steps: 8 30gradient_checkpointing: true 31group_by_length: false 32hub_model_id: error577/d09828c9-a63a-4168-9212-271729f5e2cf 33hub_repo: null 34hub_strategy: end 35hub_token: null 36learning_rate: 0.0001 37load_in_4bit: false 38load_in_8bit: false 39local_rank: null 40logging_steps: 1 41lora_alpha: 16 42lora_dropout: 0.05 43lora_fan_in_fan_out: null 44lora_model_dir: null 45lora_r: 32 46lora_target_linear: true 47lr_scheduler: cosine 48max_steps: 1000 49micro_batch_size: 2 50mlflow_experiment_name: /tmp/a0682f06fa4fb877_train_data.json 51model_type: AutoModelForCausalLM 52num_epochs: 20 53optimizer: adamw_bnb_8bit 54output_dir: miner_id_24 55pad_to_sequence_len: true 56resume_from_checkpoint: null 57s2_attention: null 58sample_packing: false 59saves_per_epoch: 1 60sequence_len: 4096 61strict: false 62tf32: false 63tokenizer_type: AutoTokenizer 64train_on_inputs: false 65trust_remote_code: true 66val_set_size: 0.02 67wandb_entity: null 68wandb_mode: online 69wandb_name: 981e1948-88b4-489c-9ec2-583892ddfa73 70wandb_project: Gradients-On-Demand 71wandb_run: your_name 72wandb_runid: 981e1948-88b4-489c-9ec2-583892ddfa73 73warmup_steps: 10 74weight_decay: 0.0 75xformers_attention: null 76

d09828c9-a63a-4168-9212-271729f5e2cf

• Loss: 2.8314

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: 10
• 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