Metadata-Version: 2.4
Name: mimo-mlx
Version: 0.1.0
Summary: MLX-native MiMo-V2.5-ASR inference package for Apple Silicon
Project-URL: Homepage, https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-INT4
Project-URL: Repository, https://github.com/XiaomiMiMo/MiMo-V2.5-ASR
Project-URL: Base Model, https://huggingface.co/XiaomiMiMo/MiMo-V2.5-ASR
Project-URL: INT4 Weights, https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-INT4
Project-URL: BF16 Weights, https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-BF16
Project-URL: F32 Weights, https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-F32
Author: Carlos Huang
License-Expression: Apache-2.0
License-File: LICENSE
Keywords: apple-silicon,asr,mimo,mlx,speech-recognition
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Requires-Dist: huggingface-hub>=1.4.0
Requires-Dist: mlx-audio>=0.3.1
Requires-Dist: mlx-lm>=0.30.5
Requires-Dist: mlx>=0.30.5
Requires-Dist: numpy>=2.0
Requires-Dist: transformers>=4.45
Description-Content-Type: text/markdown

# mimo-mlx

MLX-native MiMo-V2.5-ASR inference package for Apple Silicon.

This package contains the Python inference code only. Model weights are hosted
on Hugging Face as separate repositories:

- INT4: https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-INT4
- BF16: https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-BF16
- F32: https://huggingface.co/carloshuang1224/MiMo-V2.5-ASR-MLX-F32

The MiMo audio tokenizer is required and is downloaded from
https://huggingface.co/XiaomiMiMo/MiMo-Audio-Tokenizer when using
`download=True`.

## Install

```bash
pip install mimo-mlx
```

## Quick Start

```python
from mimo_mlx import load_asr

asr = load_asr("int4", download=True)
text = asr.transcribe("audio.wav", language="zh")
print(text)
```

Use bf16 for the validated fallback path:

```python
from mimo_mlx import load_asr

asr = load_asr("bf16", download=True)
```

If the model directories are already present locally:

```python
from mimo_mlx import load_asr

asr = load_asr("int4")
```

Expected local directory layout:

```text
models/
  MiMo-Audio-Tokenizer/
  MiMo-V2.5-ASR-int4/
  MiMo-V2.5-ASR-bf16/
```

## Notes

- The INT4 package uses MLX quantized safetensors and requires the quantized
  load path in this package.
- The default `MiMoASR()` constructor uses local bf16 paths for quality
  fallback.
- Model weights are not bundled in the PyPI package.
