mlx-MOSS-Transcribe-Diarize 4bit

4bit MLX weight-only quantized variant of vanch007/mlx-MOSS-Transcribe-Diarize.

Quantization uses MLX affine quantization with bits=4 and group_size=64 on the text backbone, excluding the Whisper audio encoder and VQ adaptor.

python -m moss_transcribe_diarize.mlx.cli /path/to/input.wav \
  --model vanch007/mlx-MOSS-Transcribe-Diarize-4bit \
  --out-dir runs/mlx_4bit_example