Audio-to-Audio
MLX
Safetensors
moonbox
mlx-audio
mossformer2
clearervoice
speech-enhancement
denoising
48khz
Instructions to use moonbox-dev/MossFormer2_SE_48K-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use moonbox-dev/MossFormer2_SE_48K-MLX with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download moonbox-dev/MossFormer2_SE_48K-MLX --local-dir MossFormer2_SE_48K-MLX
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download 6bit/config.json from moonbox-dev/MossFormer2_SE_48K-MLX: direct link, hf CLI and curl.
- Browser
- Download file 456 Bytes
-
https://huggingface.co/moonbox-dev/MossFormer2_SE_48K-MLX/resolve/main/6bit/config.json
- Command line
-
hf download hf://moonbox-dev/MossFormer2_SE_48K-MLX/6bit/config.json
-
curl -L -o config.json https://huggingface.co/moonbox-dev/MossFormer2_SE_48K-MLX/resolve/main/6bit/config.json
456 Bytes
| { | |
| "sample_rate": 48000, | |
| "win_len": 1920, | |
| "win_inc": 384, | |
| "fft_len": 1920, | |
| "win_type": "hamming", | |
| "num_mels": 60, | |
| "preemphasis": 0.97, | |
| "one_time_decode_length": 20, | |
| "decode_window": 4, | |
| "chunk_seconds": 4.0, | |
| "chunk_overlap": 0.25, | |
| "auto_chunk_threshold": 60.0, | |
| "in_channels": 180, | |
| "out_channels": 512, | |
| "out_channels_final": 961, | |
| "num_blocks": 24, | |
| "model_type": "mossformer2_se", | |
| "quantization_config": { | |
| "group_size": 64, | |
| "bits": 6 | |
| } | |
| } |