| { | |
| "epochs": 20, | |
| "batch_size": 8, | |
| "learning_rate": 1e-05, | |
| "base_dir": "/shared/nas2/ph16/toxic/outputs/classifier", | |
| "data_dir": "/shared/nas2/ph16/toxic/outputs/states/Qwen2.5-7B-Instruct", | |
| "job_name": "Qwen2.5-7B-Instruct_both/layer28", | |
| "ckpt": "", | |
| "gpu": "", | |
| "wandb": false, | |
| "overwrite": true, | |
| "hidden_sizes": [ | |
| "64" | |
| ], | |
| "random_seed": 42, | |
| "token_rule": "last", | |
| "label": "both", | |
| "n_decode": 0, | |
| "layer_id": 28, | |
| "num_classification": 2, | |
| "neg_weight": 1.0, | |
| "lens_path": "", | |
| "llm": "Qwen2.5-7B-Instruct", | |
| "use_lens": false | |
| } |