| { |
| "accelerator_kwargs": {}, |
| "adap_kl_ctrl": true, |
| "backward_batch_size": 32, |
| "batch_size": 32, |
| "cliprange": 0.2, |
| "cliprange_value": 0.2, |
| "compare_steps": 1, |
| "early_stopping": false, |
| "exp_name": "colab_kernel_launcher", |
| "forward_batch_size": null, |
| "gamma": 1, |
| "global_backward_batch_size": 32, |
| "global_batch_size": 32, |
| "gradient_accumulation_steps": 1, |
| "horizon": 10000, |
| "init_kl_coef": 0.1, |
| "is_encoder_decoder": false, |
| "is_peft_model": true, |
| "kl_penalty": "kl", |
| "lam": 0.95, |
| "learning_rate": 0.0001, |
| "log_with": null, |
| "max_grad_norm": null, |
| "mini_batch_size": 32, |
| "model_name": "gpt2", |
| "optimize_cuda_cache": true, |
| "optimize_device_cache": false, |
| "ppo_epochs": 1, |
| "project_kwargs": {}, |
| "push_to_hub_if_best_kwargs": {}, |
| "query_dataset": "imdb", |
| "ratio_threshold": 10.0, |
| "remove_unused_columns": true, |
| "reward_model": "sentiment-analysis:lvwerra/distilbert-imdb", |
| "score_clip": null, |
| "seed": 0, |
| "steps": 10, |
| "target": 0.1, |
| "target_kl": 1, |
| "task_name": null, |
| "tracker_kwargs": {}, |
| "tracker_project_name": "trl", |
| "use_score_norm": false, |
| "use_score_scaling": false, |
| "vf_coef": 0.1, |
| "whiten_rewards": false, |
| "world_size": 1 |
| } |
