See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2base_model: unsloth/Qwen2.5-Math-1.5B-Instruct 3bf16: true 4chat_template: llama3 5data_processes: 24 6dataset_prepared_path: null 7datasets: 8- data_files: 9 - 610f8682142c8104_train_data.json 10 ds_type: json 11 format: custom 12 path: /workspace/input_data/610f8682142c8104_train_data.json 13 type: 14 field_instruction: instruction 15 field_output: response 16 format: '{instruction}' 17 no_input_format: '{instruction}' 18 system_format: '{system}' 19 system_prompt: '' 20debug: null 21deepspeed: null 22device_map: auto 23do_eval: true 24early_stopping_patience: 2 25eval_batch_size: 4 26eval_max_new_tokens: 128 27eval_steps: 150 28eval_table_size: null 29evals_per_epoch: null 30flash_attention: true 31fp16: false 32fsdp: null 33fsdp_config: null 34gradient_accumulation_steps: 4 35gradient_checkpointing: false 36group_by_length: true 37hub_model_id: cimol/dafd1460-9bbc-4fe5-a329-1533e57ce948 38hub_repo: null 39hub_strategy: end 40hub_token: null 41learning_rate: 0.0002 42load_in_4bit: false 43load_in_8bit: false 44local_rank: null 45logging_steps: 50 46lora_alpha: 64 47lora_dropout: 0.05 48lora_fan_in_fan_out: null 49lora_model_dir: null 50lora_r: 32 51lora_target_linear: true 52lr_scheduler: constant 53lr_scheduler_warmup_steps: 50 54max_grad_norm: 1.0 55max_memory: 56 0: 75GB 57max_steps: 8972 58micro_batch_size: 4 59mlflow_experiment_name: /tmp/610f8682142c8104_train_data.json 60model_type: AutoModelForCausalLM 61num_epochs: 10 62optim_args: 63 adam_beta1: 0.9 64 adam_beta2: 0.95 65 adam_epsilon: 1e-5 66optimizer: adamw_bnb_8bit 67output_dir: miner_id_24 68pad_to_sequence_len: true 69resume_from_checkpoint: null 70s2_attention: null 71sample_packing: false 72save_steps: 150 73saves_per_epoch: null 74seed: 17333 75sequence_len: 512 76strict: false 77tf32: true 78tokenizer_type: AutoTokenizer 79total_train_batch_size: 16 80train_batch_size: 4 81train_on_inputs: false 82trust_remote_code: true 83val_set_size: 0.05 84wandb_entity: null 85wandb_mode: online 86wandb_name: 789b7e04-42dd-45e0-b9c3-c954ddcb3301 87wandb_project: Gradients-On-Demand 88wandb_run: your_name 89wandb_runid: 789b7e04-42dd-45e0-b9c3-c954ddcb3301 90warmup_steps: 50 91weight_decay: 0.0 92xformers_attention: null 93

dafd1460-9bbc-4fe5-a329-1533e57ce948

• Loss: 0.5346

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: 17333
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.95,adam_epsilon=1e-5
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 50
• training_steps: 8972

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