espnet2.speechlm.bin.prepare_length_stats.collect_length_stats

source

espnet2.speechlm.bin.prepare_length_stats.collect_length_stats(preprocessor, num_workers: int, spec_type: str, specifier: str) → Dict[Tuple[str, ...], int]

Collect length statistics from all worker processes.

• Parameters:
  • preprocessor – Data preprocessor
  • num_workers – Number of parallel workers
  • spec_type – Either “unregistered” or “registered”
  • specifier – The data specifier string
• Returns: Aggregated statistics dictionary