Instructions to use rueispaw/Breeze-ASR-25-mlx-q8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use rueispaw/Breeze-ASR-25-mlx-q8 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir Breeze-ASR-25-mlx-q8 rueispaw/Breeze-ASR-25-mlx-q8
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Breeze-ASR-25 (MLX, 8-bit)
MediaTek-Research/Breeze-ASR-25 的 MLX 格式 8-bit 量化版(group_size=64),供 mlx-whisper 在 Apple Silicon 上使用。
Breeze-ASR-25 是 Whisper-large-v2 針對台灣華語與華英夾雜(code-switching) 微調的 ASR 模型,原始模型與本轉換版皆為 Apache-2.0 授權。
權重取自社群轉換版 david20571015/Breeze-ASR-25-mlx-q8, 在此重新上傳以確保長期可用性。已驗證:四段測試語音(含中英夾雜、39 秒 長音訊)的完整轉寫輸出,與官方權重在 PyTorch 下的結果逐字元相同。
使用
import mlx_whisper
result = mlx_whisper.transcribe(
"audio.wav",
path_or_hf_repo="rueispaw/Breeze-ASR-25-mlx-q8",
language="zh",
task="transcribe",
)
print(result["text"])
- Downloads last month
- 18
Hardware compatibility
Log In to add your hardware
Quantized
Model tree for rueispaw/Breeze-ASR-25-mlx-q8
Base model
openai/whisper-large-v2 Finetuned
MediaTek-Research/Breeze-ASR-25