README.md

---
language: vi
license: mit
library_name: transformers
pipeline_tag: automatic-speech-recognition
base_model: openai/whisper-large-v3
tags: [automatic-speech-recognition, whisper, vietnamese, buzzasr]
datasets: [google/fleurs]
metrics: [cer, wer]
---

# BuzzASR — Vietnamese

A monolingual automatic speech recognition model for **Vietnamese**, fine-tuned from
[openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3). Part of **BuzzASR**,
a suite of 102 language-specialized ASR models (Findings of EMNLP 2026).

This model uses **simple fine-tuning (Whisper's tokenizer, ASR fine-tuning only)**.

## Results (normalized CER / WER, %)

| Test set | CER | WER | Whisper-large-v3 (zero-shot) CER |
|---|---|---|---|
| FLEURS | 3.73 | 7.26 | 3.9 |
| Combined | 5.38 | 10.68 | 3.9 |

~0.7x CER reduction over Whisper zero-shot on the combined test set.

## Usage

```python
import torch, torchaudio
from transformers import WhisperForConditionalGeneration, WhisperProcessor

model = WhisperForConditionalGeneration.from_pretrained("BuzzASR/vietnamese", torch_dtype=torch.float16).to("cuda").eval()
proc  = WhisperProcessor.from_pretrained("BuzzASR/vietnamese")

wav, sr = torchaudio.load("audio.wav")           # 16 kHz mono
feats = proc(wav[0], sampling_rate=16000, return_tensors="pt").input_features.to("cuda").half()
ids = model.generate(feats, num_beams=1, no_repeat_ngram_size=3, repetition_penalty=1.2)
print(proc.batch_decode(ids, skip_special_tokens=True)[0])
```
The language/task prompt is baked into the generation config, so no `language=` argument is needed.

## Training data
[FLEURS](https://huggingface.co/datasets/google/fleurs), capped per the paper. Text-only data from the **Goldfish** corpus (Chang et al., 2026).

## Limitations
Monolingual (Vietnamese only). Evaluated on FLEURS / Common Voice test splits; other domains or dialects may differ.

## Citation
Project page: https://lemn-lab.github.io/buzzasr-docs/