| { | |
| "model_name": "microsoft/deberta-v3-small", | |
| "max_len": 256, | |
| "num_labels": 955, | |
| "min_heading_freq": 8, | |
| "other_label": "Other", | |
| "text_format": "[DEPT] {department} [SEP] {question_clean}", | |
| "train_rows": 198372, | |
| "val_rows": 22042, | |
| "seed": 42, | |
| "learning_rate": 2e-05, | |
| "batch_size": 16, | |
| "epochs": 4, | |
| "tokenizer_fast": false, | |
| "tokenizer_class": "DebertaV2Tokenizer", | |
| "early_stopping_patience": 1, | |
| "early_stopping_threshold": 0.0, | |
| "best_model_metric": "eval_top1_accuracy" | |
| } |