---
language: en
tags:
- nemo
- asr
- children-speech
- parakeet
- tdt
license: cc-by-4.0
base_model: nvidia/parakeet-tdt-0.6b-v2
---

# Parakeet TDT 0.6B v2 - Fine-tuned on Children's Speech

Fine-tuned from [nvidia/parakeet-tdt-0.6b-v2](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2)
on children's speech data from the
[DrivenData ASR competition](https://www.drivendata.org/competitions/288/).

## Training config

- Mode: full
- Epochs: 10
- Batch size: 12 (accumulate: 2)
- Learning rate: 3e-05
- Precision: bf16-mixed
- Speed perturbation: True

## Usage

```python
import nemo.collections.asr as nemo_asr

model = nemo_asr.models.ASRModel.restore_from("best_model.nemo")
hypotheses = model.transcribe(["audio.flac"])
print(hypotheses[0].text)
```