---
language:
- ta
task_categories:
- automatic-speech-recognition
tags:
- audio
- speech
- nemo
- tarred
---

# NeMo Tarred Dataset

Generated from `MRTS`.

Train rows: 76080 · Test rows: 736
Shards: 16 · Codec: wav · Sample rate: 16000 Hz mono
Primary text: `text` · target_lang: `ta-IN`

```yaml
is_tarred: true
tarred_audio_filepaths: .../audio__OP_0..15_CL_.tar
manifest_filepath: .../train_manifest.json
```