{
  "encoder_model_id": "KeisukeMiyamoto/zipformer-cpt",
  "decoder_model_id": "KeisukeMiyamoto/lfm2.5-230m-cpt",
  "sampling_rate": 16000,
  "output_dir": "outputs/training",
  "resume_checkpoint": null,
  "dataset_cache_dir": "outputs/datasets/lambda-audio-courpus",
  "inference_samples_path": "data/data.parquet",
  "max_audio_seconds": 30.0,
  "max_new_tokens": 256,
  "bucket_buffer_size": 512,
  "num_workers": 12,
  "prefetch_factor": 4,
  "max_batch_padded_audio_seconds": 1800.0,
  "gradient_accumulation_steps": 1,
  "max_steps": 64000,
  "unfreeze_step": 4000,
  "adapter_learning_rate": 0.00005,
  "backbone_learning_rate": 0.00005,
  "warmup_steps": 2000,
  "min_learning_rate_ratio": 0.2,
  "weight_decay": 0.01,
  "max_grad_norm": 1.0,
  "log_steps": 10,
  "evaluation_steps": 2000,
  "save_steps": 500,
  "early_stopping_patience": 5,
  "early_stopping_min_delta": 0.01,
  "seed": 42
}