| # pytest: disable | |
| # Model arguments | |
| model_name_or_path: AMD-OLMo-1B-dpo | |
| torch_dtype: null | |
| use_flash_attention_2: false | |
| chat_template: "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n' + message['content'] }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}" | |
| # Data training arguments | |
| # For definitions, see: src/h4/training/config.py | |
| dataset_mixer: | |
| csarron/argilla-ultrafeedback-binarized-preferences-cleaned: 1.0 | |
| dataset_splits: | |
| - train | |
| - test | |
| preprocessing_num_workers: 16 | |
| # DPOTrainer arguments | |
| bf16: true | |
| beta: 0.01 | |
| do_eval: true | |
| evaluation_strategy: steps | |
| eval_steps: 100 | |
| gradient_accumulation_steps: 2 | |
| gradient_checkpointing: true | |
| gradient_checkpointing_kwargs: | |
| use_reentrant: False | |
| hub_model_id: AMD-OLMo-1B-dpo | |
| learning_rate: 5.0e-5 | |
| log_level: info | |
| logging_steps: 10 | |
| lr_scheduler_type: cosine | |
| max_length: 1024 | |
| max_prompt_length: 512 | |
| num_train_epochs: 3 | |
| optim: adamw_torch | |
| output_dir: data/AMD-OLMo-1B-dpo | |
| per_device_train_batch_size: 8 | |
| per_device_eval_batch_size: 8 | |
| push_to_hub: false | |
| save_strategy: "steps" | |
| save_steps: 100 | |
| save_total_limit: 1 | |
| seed: 42 | |
| warmup_ratio: 0.1 |