span_training_extra_config.json

{
  "saved_at": "2026-05-23T11:10:39",
  "best_epoch": 8,
  "best_validation_score_char_f1": 0.9950298210735586,
  "best_validation_config": {
    "threshold": 0.5,
    "min_span_chars": 1,
    "min_span_tokens": 1,
    "merge_gap_chars": 1,
    "strip_predicted_span_whitespace": true,
    "drop_spans_without_alnum": true,
    "score_name": "sum_non_O_probability"
  },
  "training_config": {
    "base_model_name": "microsoft/deberta-v3-small",
    "max_length": 512,
    "span_labels": [
      "O",
      "tool_output_conflict",
      "overgeneration",
      "missing_tool_action_recommendation"
    ],
    "label2id": {
      "O": 0,
      "tool_output_conflict": 1,
      "overgeneration": 2,
      "missing_tool_action_recommendation": 3
    },
    "id2label": {
      "0": "O",
      "1": "tool_output_conflict",
      "2": "overgeneration",
      "3": "missing_tool_action_recommendation"
    },
    "micro_batch_size": 4,
    "gradient_accumulation_steps": 4,
    "effective_batch_size": 16,
    "num_epochs": 8,
    "patience": 3,
    "learning_rate": 2e-05,
    "weight_decay": 0.01,
    "warmup_ratio": 0.1,
    "warmup_steps": 90,
    "total_update_steps": 904,
    "max_grad_norm": 1.0,
    "use_amp": false,
    "force_fp32_training": true,
    "amp_disabled_reason": "Previous run failed with GradScaler ValueError: Attempting to unscale FP16 gradients.",
    "use_class_weights": true,
    "class_weights": {
      "O": 0.14653727412223816,
      "tool_output_conflict": 1.7199265956878662,
      "overgeneration": 1.2583367824554443,
      "missing_tool_action_recommendation": 0.8751992583274841
    },
    "threshold_grid": [
      0.05,
      0.08,
      0.1,
      0.12,
      0.15,
      0.18,
      0.2,
      0.25,
      0.3,
      0.35,
      0.4,
      0.45,
      0.5,
      0.55,
      0.6,
      0.65,
      0.7,
      0.75,
      0.8,
      0.85,
      0.9,
      0.95
    ],
    "min_span_chars_grid": [
      1,
      2,
      3,
      5,
      8,
      12
    ],
    "min_span_tokens_grid": [
      1,
      2
    ],
    "merge_gap_chars": 1
  },
  "dataset_report": {
    "tokenizer_class": "DebertaV2Tokenizer",
    "model_max_length": 1000000000000000019884624838656,
    "max_length": 512,
    "train_windows": 1808,
    "validation_windows": 229,
    "test_windows": 235,
    "train_supervised_token_count": 246653,
    "validation_supervised_token_count": 29627,
    "test_supervised_token_count": 33033,
    "train_positive_token_count": 10688,
    "validation_positive_token_count": 1441,
    "test_positive_token_count": 1459,
    "train_token_label_counts": {
      "O": 235965,
      "overgeneration": 3200,
      "missing_tool_action_recommendation": 6615,
      "tool_output_conflict": 873
    },
    "validation_token_label_counts": {
      "O": 28186,
      "overgeneration": 428,
      "tool_output_conflict": 121,
      "missing_tool_action_recommendation": 892
    },
    "test_token_label_counts": {
      "O": 31574,
      "missing_tool_action_recommendation": 925,
      "tool_output_conflict": 128,
      "overgeneration": 406
    },
    "class_weights": {
      "O": 0.14653727412223816,
      "tool_output_conflict": 1.7199265956878662,
      "overgeneration": 1.2583367824554443,
      "missing_tool_action_recommendation": 0.8751992583274841
    }
  }
}