| { |
| "passages_path": "data/passages.json", |
| "train_path": "data/splits/train.json", |
| "val_path": "data/splits/val.json", |
| "out_dir": "inferences/dpr_ckpt_production", |
| "backbone": "vinai/phobert-base", |
| "init_model_dir": "inferences/dpr_ckpt_1", |
| "init_weights": "model_last.pt", |
| "epochs": 40, |
| "batch_size": 256, |
| "passage_batch_size": 64, |
| "lr": 0.001, |
| "weight_decay": 0.01, |
| "warmup_ratio": 0.08, |
| "temperature": 0.05, |
| "label_smoothing": 0.0, |
| "max_q_len": 128, |
| "max_p_len": 512, |
| "grad_accum_steps": 1, |
| "max_grad_norm": 1.0, |
| "device": "mps", |
| "seed": 42, |
| "num_threads": 4, |
| "gradient_checkpointing": false, |
| "freeze_encoder": true, |
| "cache_encoder_features": true, |
| "architecture": "DualEncoder", |
| "pooling": "mean", |
| "projection_dim": 768, |
| "training_mode": "full_softmax_all_passages", |
| "passage_count": 91, |
| "metrics": { |
| "loss": 2.9483868518989245, |
| "mrr": 0.684004583772982, |
| "recall@1": 0.5469061876247505, |
| "recall@3": 0.7924151696606786, |
| "recall@5": 0.8762475049900199, |
| "recall@10": 0.9261477045908184, |
| "recall@20": 0.9341317365269461, |
| "recall@50": 0.9640718562874252 |
| } |
| } |
