---
language:
  - am
license: apache-2.0
library_name: transformers
pipeline_tag: automatic-speech-recognition
base_model: Qwen/Qwen3-ASR-0.6B
tags:
  - qwen3-asr
  - automatic-speech-recognition
  - amharic
  - speech
---

# Qwen3-ASR-0.6B Amharic - Curated

An Amharic automatic speech recognition checkpoint derived from
[Qwen3-ASR-0.6B](https://huggingface.co/Qwen/Qwen3-ASR-0.6B).

This variant is tuned on a curated Amharic speech-text corpus and is the
recommended release when performance on the curated-source evaluation is the
priority.

## Benchmark

Results use fixed held-out evaluation sets and identical deterministic
generation settings (`language=None`, `max_new_tokens=512`, batch size 1).
Lower WER and CER are better.

| Evaluation | Curated variant | Broader mixed variant |
|---|---:|---:|
| Curated-source heldout WER | **36.99%** | 37.46% |
| Curated-source heldout CER | **15.71%** | 16.18% |
| WAXAL WER | 37.56% | **36.57%** |
| WAXAL CER | 16.45% | **15.57%** |

The curated variant is stronger on the curated-source heldout. The broader
mixed variant generalizes better to WAXAL in this evaluation.

## Live Colab Demo

[Open the Gradio demo notebook](https://huggingface.co/b1n1yam/qwen3-asr-0.6b-amharic-gold-silver/blob/main/colab_demo.ipynb),
then open it in Google Colab and run the cells in order. The final cell creates
a temporary public Gradio link while the notebook remains connected.

## Usage

```bash
pip install -U qwen-asr
```

```python
import torch
from qwen_asr import Qwen3ASRModel

model = Qwen3ASRModel.from_pretrained(
    "b1n1yam/qwen3-asr-0.6b-amharic-gold-silver",
    dtype=torch.bfloat16,
    device_map="cuda:0",
    max_inference_batch_size=8,
    max_new_tokens=512,
)

result = model.transcribe(
    audio="path/to/audio.wav",
    language=None,
)

print(result[0].text)
```

Audio should contain clear Amharic speech. Keep `language=None`; this is the
setting used for the reported benchmarks.

## Intended Use

- Amharic speech transcription
- Research and evaluation of Amharic ASR
- Further domain adaptation with appropriately licensed data

## Limitations

- Performance varies by recording conditions, speaker, dialect, domain, and
  audio duration.
- The model can omit words or generate incorrect text. Human review is
  required for consequential uses.
- The benchmark covers a limited set of held-out corpora and should not be
  interpreted as universal Amharic ASR performance.
- This release is not designed for speaker identification, translation, or
  timestamp alignment.

## Training Data Disclosure

The model was adapted with a privately maintained, curated Amharic
speech-text corpus. Source manifests and row-level metadata are not included
in this repository.

## License and Attribution

This model is released under the Apache-2.0 license, following the upstream
Qwen3-ASR-0.6B license. Please also review the
[upstream model card](https://huggingface.co/Qwen/Qwen3-ASR-0.6B).