training_config.json

{
  "squad_v2": false,
  "model_checkpoint": "albert-base-v2",
  "max_length": 512,
  "doc_stride": 64,
  "n_best_size": 100,
  "max_answer_length": 64,
  "use_data_augmentation": true,
  "augment_probability": 0.15,
  "context_dropout": 0.05,
  "question_paraphrasing": true,
  "negative_sampling": true,
  "batch_size": 28,
  "num_epochs": 6,
  "learning_rate": 6e-05,
  "weight_decay": 0.005,
  "warmup_ratio": 0.08,
  "gradient_accumulation_steps": 2,
  "max_grad_norm": 0.5,
  "optimizer_type": "adamw",
  "optimizer_betas": [
    0.9,
    0.98
  ],
  "optimizer_eps": 1e-07,
  "scheduler_type": "cosine_with_restarts",
  "scheduler_power": 0.5,
  "scheduler_eta_min": 5e-07,
  "scheduler_num_cycles": 0.5,
  "early_stopping_patience": 4,
  "early_stopping_threshold": 0.0002,
  "early_stopping_metric": "f1",
  "log_interval": 50,
  "eval_steps": null,
  "save_steps": null,
  "save_total_limit": 3,
  "wandb_project": "question-answering-enhanced",
  "wandb_entity": null,
  "use_wandb": true,
  "wandb_tags": [
    "question-answering",
    "squad",
    "multi-gpu"
  ],
  "push_to_hub": true,
  "hub_model_id": "HariomSahu/albert-squad-qa",
  "hub_private": false,
  "hub_model_name_max_length": 50,
  "hub_versioning_strategy": "single_repo_versions",
  "hub_base_model_name": "albert-squad-qa",
  "use_label_smoothing": true,
  "label_smoothing_factor": 0.1,
  "use_focal_loss": false,
  "focal_loss_alpha": 0.25,
  "focal_loss_gamma": 2.0,
  "use_curriculum_learning": true,
  "curriculum_strategy": "length_based",
  "dropout_rate": 0.1,
  "attention_dropout": 0.1,
  "use_mixup": false,
  "mixup_alpha": 0.2,
  "seed": 42,
  "dataloader_num_workers": 0,
  "dataloader_pin_memory": true
}