See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2base_model: bigscience/bloomz-560m 3bf16: true 4chat_template: llama3 5dataset_prepared_path: null 6datasets: 7- data_files: 8 - 25381d550887dd6e_train_data.json 9 ds_type: json 10 format: custom 11 path: /workspace/input_data/25381d550887dd6e_train_data.json 12 type: 13 field_input: operators 14 field_instruction: question_text 15 field_output: decomposition 16 format: '{instruction} {input}' 17 no_input_format: '{instruction}' 18 system_format: '{system}' 19 system_prompt: '' 20debug: null 21deepspeed: null 22early_stopping_patience: 2 23eval_max_new_tokens: 128 24eval_steps: 100 25eval_table_size: null 26flash_attention: false 27fp16: null 28fsdp: null 29fsdp_config: null 30gradient_accumulation_steps: 8 31gradient_checkpointing: true 32group_by_length: false 33hub_model_id: Alphatao/a8b12708-2aae-43c9-9def-7df8e7127f74 34hub_repo: null 35hub_strategy: checkpoint 36hub_token: null 37learning_rate: 0.0002 38load_best_model_at_end: true 39load_in_4bit: false 40load_in_8bit: false 41local_rank: null 42logging_steps: 1 43lora_alpha: 32 44lora_dropout: 0.05 45lora_fan_in_fan_out: null 46lora_model_dir: null 47lora_r: 16 48lora_target_linear: true 49lora_target_modules: null 50lr_scheduler: cosine 51max_grad_norm: 1.0 52max_steps: 6528 53micro_batch_size: 4 54mlflow_experiment_name: /tmp/25381d550887dd6e_train_data.json 55model_type: AutoModelForCausalLM 56num_epochs: 2 57optimizer: adamw_bnb_8bit 58output_dir: miner_id_24 59pad_to_sequence_len: true 60resume_from_checkpoint: null 61s2_attention: null 62sample_packing: false 63save_steps: 100 64sequence_len: 1024 65strict: false 66tf32: true 67tokenizer_type: AutoTokenizer 68train_on_inputs: false 69trust_remote_code: true 70val_set_size: 0.04 71wandb_entity: null 72wandb_mode: online 73wandb_name: d7f41241-8e43-47ec-9de5-2ca5bb717ff5 74wandb_project: Gradients-On-Demand 75wandb_run: your_name 76wandb_runid: d7f41241-8e43-47ec-9de5-2ca5bb717ff5 77warmup_steps: 10 78weight_decay: 0.0 79xformers_attention: null 80

a8b12708-2aae-43c9-9def-7df8e7127f74

• Loss: 0.2028

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0002
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• 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: 3591

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