---
configs:
- config_name: eval-control-record-female-north
  data_files:
  - split: test
    path: eval-control-record-female-north/test-00000-of-00001.parquet
- config_name: eval-synth-400
  data_files:
  - split: test
    path: eval-synth-400/test-00000-of-00001.parquet
- config_name: non-command
  data_files:
  - split: test
    path: non-command/test-00000-of-00001.parquet
---

# phongps2/eval-speech

Speech evaluation dataset with multiple benchmark configs.

## Available Configs

- `eval-control-record-female-north`: control-command evaluation set, split `test`
- `eval-synth-400`: synthetic evaluation set, split `test`
- `non-command`: non-command / noise-style evaluation set, split `test`

## Notes

- Files are stored as parquet for dataset viewer compatibility.
- The `eval-control-record-female-north` config includes an audio feature for direct playback in the dataset viewer.