{ "task": "constraint_dimension", "epoch_stats": [ { "epoch": 1, "train_loss": 1.0786596297055173 }, { "epoch": 2, "train_loss": 0.49559883196265825 }, { "epoch": 3, "train_loss": 0.36042651461985464 }, { "epoch": 4, "train_loss": 0.27737118249450576 }, { "epoch": 5, "train_loss": 0.20896544052678293 }, { "epoch": 6, "train_loss": 0.15534398090000953 }, { "epoch": 7, "train_loss": 0.11773543272898254 }, { "epoch": 8, "train_loss": 0.07906012590035061 }, { "epoch": 9, "train_loss": 0.05605952438313094 }, { "epoch": 10, "train_loss": 0.03452488826215441 } ], "training_summary": { "actual_epochs": 10, "best_epoch": 10, "early_stopped": false, "backbone_model_name": "microsoft/deberta-v3-base", "tokenizer_name": "microsoft/deberta-v3-base" }, "calibration": { "method": "temperature_grid_search", "rows": 6000, "temperature": 2.0, "loss": 0.5294126563569427 }, "hf_model_dir": "/nvme/CognitiveMemoryLayer/packages/models/trained_models/constraint_dimension_hf" }