Instructions to use younghan-meta/Supertonic-ExecuTorch-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use younghan-meta/Supertonic-ExecuTorch-MLX with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir Supertonic-ExecuTorch-MLX younghan-meta/Supertonic-ExecuTorch-MLX
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Supertonic 3 ExecuTorch MLX Export
This repository contains a ready-to-run ExecuTorch PTE export of
Supertone/supertonic-3 for
Apple silicon Macs using the ExecuTorch MLX backend.
Artifact
| File | Format | Size | SHA-256 |
|---|---|---|---|
model.pte |
ExecuTorch PTE | 198,747,136 bytes | 8aa17b185db9c4b18c8bebd49414337342a0499985e8218d6c30f618c8227432 |
The export was prepared from:
- Source model:
Supertone/supertonic-3 - Source revision:
3cadd1ee6394adea1bd021217a0e650ede09a323 - ExecuTorch export revision:
20ad5ee43ff53804030899d621590af3daadda53 - Target platform: macOS arm64
- Backend: ExecuTorch MLX
manifest.json contains the same provenance and integrity metadata in a
machine-readable form.
Runtime requirements
This repository intentionally does not redistribute:
- the native
supertonic_runnerexecutable; mlx.metallib;- the original Supertonic configuration and text-processing assets;
- voice-style JSON files.
Build the native runner from the pinned ExecuTorch source and obtain the model assets and a voice style from the pinned upstream Supertonic repository. The Meta Model Cookbook realtime voice-agent recipe automates the public artifact download and integrity checks while keeping the native runner local.
License and use restrictions
This PTE is a converted/modified form of the Supertonic model. It remains
subject to the BigScience Open RAIL-M License, including the use-based
restrictions in Attachment A. Read LICENSE before downloading, using, or
redistributing the model.
The conversion changes the serialization and execution format for the
ExecuTorch MLX runtime; it does not retrain or fine-tune the model. See
NOTICE.md for attribution and modification details.
Verification
shasum -a 256 model.pte
Expected result:
8aa17b185db9c4b18c8bebd49414337342a0499985e8218d6c30f618c8227432 model.pte
For reproducible applications, download this artifact by an immutable Hugging
Face repository commit rather than by the mutable main branch, then verify
both the exact byte size and SHA-256 above.
- Downloads last month
- -
Model tree for younghan-meta/Supertonic-ExecuTorch-MLX
Base model
Supertone/supertonic-3