training_config.json

{
  "model": {
    "backbone": "microsoft/deberta-v3-xsmall",
    "char_embed_dim": 50,
    "char_vocab_size": 256,
    "char_cnn_filters": [
      50,
      50,
      50
    ],
    "char_cnn_widths": [
      3,
      4,
      5
    ],
    "max_char_len": 20,
    "dropout": 0.1
  },
  "data": {
    "max_seq_len": 256,
    "val_ratio": 0.1,
    "test_ratio": 0.1,
    "seed": 42,
    "oversample_tiers": [
      1
    ],
    "oversample_factor": 3
  },
  "training": {
    "epochs": 10,
    "batch_size": 8,
    "gradient_accumulation_steps": 4,
    "lr_backbone": 1e-05,
    "lr_head": 0.001,
    "lr_scheduler_type": "cosine",
    "warmup_steps": 500,
    "weight_decay": 0.01,
    "eval_strategy": "epoch",
    "save_strategy": "epoch",
    "metric_for_best_model": "overall_f1",
    "logging_steps": 50,
    "dataloader_num_workers": 4,
    "save_total_limit": 3,
    "output_dir": "/home/ubuntu/datafog-labs/pii-ner-v1/runs/2026-02-06_v1.3",
    "run_name": "pii-ner-v1.3-h100-2026-02-06",
    "freeze_backbone_after_epoch": 3,
    "tier_weights": {
      "1": 3.0,
      "2": 2.0,
      "3": 1.5,
      "4": 1.0
    },
    "tier_weights_after_epoch_2": {
      "1": 2.0,
      "2": 1.5,
      "3": 1.25,
      "4": 1.0
    }
  },
  "wandb": {
    "enabled": true,
    "project": "datafog-pii-ner"
  }
}