training_config.json

{
  "task": "B",
  "num_labels": 11,
  "base_model": "microsoft/codebert-base",
  "stylometric_dim": 32,
  "num_gpad_layers": 1,
  "entropy_alpha": 0.15,
  "max_length": 256,
  "use_stylometric": true,
  "classifier_hidden_dim": 384,
  "num_epochs": 3,
  "train_batch_size": 8,
  "eval_batch_size": 16,
  "learning_rate": 2e-05,
  "weight_decay": 0.01,
  "warmup_ratio": 0.1,
  "gradient_accumulation_steps": 1,
  "use_mixed_precision": true,
  "use_gradient_checkpointing": true,
  "val_split": 0.2,
  "sample_size": null,
  "output_dir": "/kaggle/working/results_taskB_full-smpl",
  "log_interval": 100,
  "early_stopping_patience": 3,
  "seed": 42,
  "push_to_hub": true,
  "hub_model_id": "ranjan56cse/gpad-v2-full-taskB-smpl",
  "hub_private": false
}