Automatic Speech Recognition
MLX
Safetensors
Chinese
English
whisper
mlx-whisper
speech-to-text
traditional-chinese
taiwan
quantized
apple-silicon
5-bit precision
5-bit
Instructions to use MarkChen1214/Breeze-ASR-25-MLX-5bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use MarkChen1214/Breeze-ASR-25-MLX-5bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir Breeze-ASR-25-MLX-5bit MarkChen1214/Breeze-ASR-25-MLX-5bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
File size: 333 Bytes
f38281b | 1 2 3 4 5 6 7 8 | {
"name": "uniform-q5",
"role": "in-house uniform control",
"note": "Affine 5-bit, group 64, applied to every quantizable module. No external Breeze checkpoint exists at this precision except Q8 (rueispaw), which this reproduces as a cross-check.",
"default": {"bits": 5, "group_size": 64, "mode": "affine"},
"rules": []
}
|