| { |
| "dataset_name": "OpenAssistant/oasst1", |
| "model_name": "meta-llama/Meta-Llama-3-8B", |
| "dataset_split": "", |
| "subsample_percentage": 1, |
| "input_layer": 3, |
| "freeze_layer": false, |
| "use_lora": "True", |
| "use_quantization": "True", |
| "concat": "True", |
| "use_softprompt": "True", |
| "seed": 42, |
| "batch_size": 8, |
| "fold": 0, |
| "train_epochs": 2, |
| "max_length": 10000, |
| "learning_rate": 5e-05, |
| "weight_decay": 0.1, |
| "min_lr_ratio": 0.7, |
| "noise_factor": 0.0, |
| "use_redistribution": "True", |
| "redistribution_type": "scalar", |
| "sigma_left": 4.0, |
| "sigma_right": 5.5, |
| "redistribution_dk": 32, |
| "lr_scheduler_type": "cosine_with_min_lr", |
| "grid_search": false, |
| "gradient_acum_steps": 1, |
| "gradient_checkpointing": true, |
| "logging_steps": 50, |
| "train_samples": 0, |
| "freeze": 0, |
| "mode": "train", |
| "fp_dropout": "0.1,0.3", |
| "fixations_model_version": "2", |
| "features_used": "1,1,1,1,1" |
| } |
