{ "config": { "epochs": 3, "batch_size": 4, "gradient_accumulation": 4, "learning_rate": 3e-05, "freeze_embeddings": true, "warmup_ratio": 0.1, "weight_decay": 0.01, "label_smoothing": 0.05, "patience": 2, "bf16": true, "layer_lr_decay": 0.95, "seed": 42, "lora": true, "gradient_checkpointing": true }, "history": [ { "epoch": 1, "train_loss": 1.097389962043493, "validation": { "support": 132, "positive_support": 57, "precision": 0.8548387096774194, "recall": 0.9298245614035088, "f1": 0.8907563025210085, "macro_f1": 0.9005505650536076, "average_precision": 0.9244883870094944, "confusion_matrix": [ [ 66, 9 ], [ 4, 53 ] ], "coverage": 0.946969696969697, "review_rate": 0.05303030303030303, "review_count": 7, "redundant_count": 50, "not_redundant_count": 75, "selective_precision": 0.9, "negative_predictive_value": 0.9066666666666666 } }, { "epoch": 2, "train_loss": 0.9638782722815391, "validation": { "support": 132, "positive_support": 57, "precision": 0.8548387096774194, "recall": 0.9298245614035088, "f1": 0.8907563025210085, "macro_f1": 0.9005505650536076, "average_precision": 0.924562735648003, "confusion_matrix": [ [ 66, 9 ], [ 4, 53 ] ], "coverage": 0.946969696969697, "review_rate": 0.05303030303030303, "review_count": 7, "redundant_count": 50, "not_redundant_count": 75, "selective_precision": 0.9, "negative_predictive_value": 0.9066666666666666 } }, { "epoch": 3, "train_loss": 0.9364760390693142, "validation": { "support": 132, "positive_support": 57, "precision": 0.8548387096774194, "recall": 0.9298245614035088, "f1": 0.8907563025210085, "macro_f1": 0.9005505650536076, "average_precision": 0.92445966411636, "confusion_matrix": [ [ 66, 9 ], [ 4, 53 ] ], "coverage": 0.946969696969697, "review_rate": 0.05303030303030303, "review_count": 7, "redundant_count": 50, "not_redundant_count": 75, "selective_precision": 0.9, "negative_predictive_value": 0.9066666666666666 } } ], "selected_epoch": 2, "selected_parameter_delta_l2": 0.010931980796158314, "optimizer_steps": 141, "pruned": false, "policy": { "binary_threshold": 0.39322978258132935, "thresholds": { "low": 0.9576294422149658, "high": 0.9850045442581177, "positive_enabled": true, "negative_enabled": true }, "selection_split": "validation", "precision_target": 0.9, "negative_predictive_value_target": 0.9, "minimum_support": 5, "decision_policy_version": "0.1.0" }, "training_pairs": 744, "validation_pairs": 132, "trainable_parameters": 297219, "total_parameters": 279108870, "duration_seconds": 720.2784937520046, "environment": { "python": "3.11.14", "platform": "Linux-7.1.8+deb13-amd64-x86_64-with-glibc2.41", "code_commit": "07cfdfd966b52a81698851cd641b56ff63432d1e", "source_sha256": { "cli.py": "8462dd1c90e06165826a5053c740c2e4379bc4db1a3fafacb38eccc866f007ab", "decision.py": "ce69d18701fb687f0fd94217a121f7ebdd29e9397c0c7a18c899dcede96f5e2a", "calibration.py": "d999744c3ed24167540740e010d98deb96ab23e01570f2de24d2e12d8758a044", "pipeline.py": "780d3fb9e827ab65a9aa21ec09eaecf1f4289fb2492c23a006659f08f0384933", "encoder.py": "1536cb1330fe00f6c30e41ee0ec393f6fb964da71c7e70a0448e286d7493a6c1", "nli.py": "f74b06c83c66bb6d6f500b01ee9831f56422e8aad44e2355bff48ab1c1b27558", "research_scores.py": "9f235eb96b44db8fc48c1ce62a6f048d54e552aca70eeee296b977b2330a4d76", "pseudo.py": "b974fd870e6bf614131d097fa7f49aba77696d3b29f164a3ea1c094d7383ef74", "data.py": "5b8143e617f9166f13159f96081b9048b24fe6d80289b181fc165f1c307b7dde", "experiment.py": "fd3468655604e01ea8867e0e5aec266281239f5835d290e929728c1af2922e68", "score_graph.py": "252d95a1ba60fe26cfab1e9e57eb47b51ddf04d7b706ea52ced1f5376bd0f9ed", "domain_adaptation.py": "cab4001a9da57af5aad758d43ffcb1cb76c5cc6c97eac2c7a6c92f45a076ed8f", "training.py": "2a375b311b3a9e3e5d773e0b0f974ff859dd4f232d639c8176302128142aa8c0", "__main__.py": "b466d3224c1cfa96b82f891d496e36e7bd4ad1eb70fcc530b45c29ce62d3f9c4", "cross_encoder.py": "2fd31598b2ae3e039ffcbab908c1c6b190aa6aba74d117f5658fc2fde89b5437", "schemas.py": "361128ca989504938a2bf8977cbf0b1be49bee9296c5da4903b05736e149535c", "evaluation.py": "4880d0ee9fcf6fc15163204b2edcf4856dc06dfe5aaa3930a9dca236b16d0fd0", "__init__.py": "8f3ef65a08808435e85992bcb80379e0a060dab973ffbe2b5e193bc924766681", "v03_data.py": "989e17936dcc772dc414d6ff918929b4dd961ee279e4f337bec82579844f8479", "dataset.py": "bf9681d9a0ca6a1e283b7423d4eb367c478cffd8c9adfab5979e64b825f7648a" }, "deterministic_warn_only": true, "cublas_workspace_config": ":4096:8", "gpu": "NVIDIA RTX A1000 6GB Laptop GPU", "packages": { "torch": "2.6.0", "sentence-transformers": "5.1.2", "transformers": "4.57.1", "numpy": "2.3.4", "scikit-learn": "1.7.2" } }, "objective": "multitask directional NLI CE and symmetric redundancy BCE", "score": "minimum bidirectional entailment softmax; uncalibrated" }