{
  "architecture": "hierarchical",
  "model_name": "Qwen/Qwen3-0.6B-Base",
  "model_dtype": "auto",
  "hf_dataset": "Praveenrajus/jev-bench",
  "dataset_revision": "main",
  "processed_data_dir": "jev_bench_processed",
  "force_reprocess": false,
  "surface_seed": 42,
  "configs": null,
  "use_human_soft_labels": false,
  "quick_run": false,
  "balanced_train_sampling": true,
  "balance_field": "source_config",
  "train_samples_per_epoch": null,
  "max_train_examples": null,
  "max_val_examples": null,
  "max_test_examples": null,
  "epochs": 1,
  "batch_size": 16,
  "grad_accum_steps": 1,
  "gradient_checkpointing": true,
  "learning_rate": 1e-05,
  "weight_decay": 0.01,
  "warmup_ratio": 0.06,
  "max_grad_norm": 1.0,
  "randomize_candidate_order": true,
  "eval_every": 1000,
  "eval_max_batches": null,
  "output_dir": "checkpoints_jev_hierarchical-20",
  "seed": 42,
  "log_every": 25,
  "num_workers": 1,
  "wandb_project": "variable-decision-jev-bench",
  "wandb_entity": null,
  "run_name": "jev-bench-branching-20",
  "use_wandb": true,
  "max_length": 512,
  "max_label_length": 96,
  "label_layers": 2,
  "label_heads": 8,
  "label_ff_mult": 4,
  "candidate_layers": 0,
  "candidate_heads": 8,
  "candidate_ff_mult": 4,
  "dropout": 0.0,
  "parameter_summary": {
    "total": 621347841,
    "trainable": 621347841,
    "frozen": 0,
    "trainable_percent": 100.0
  },
  "checkpoint": {
    "name": "best_val",
    "score": 0.7079759862778731,
    "loss": 0.8373165361994181,
    "step": 8373
  }
}