Remove 61 per-shard train metadata parquets (superseded by metadata/train.parquet)

Add consolidated metadata/train.parquet; update README

Remove pure non-speech segments from test/validation (audio + metadata); update README

Remove legacy/incomplete CSV metadata (replaced by parquet)

Update README: document parquet metadata, fix citation, add changelog

Replace incomplete CSV metadata with complete per-shard parquet metadata (regenerated from source parquets; duration as float64)

Delete gigaspeech

Upload folder using huggingface_hub

Update README.md

Upload folder using huggingface_hub

Upload script/process_from_parquet.py with huggingface_hub

Upload script/gigaspeech.py with huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload gigaspeech-parquet/validation.parquet with huggingface_hub

initial commit