software-mansion/react-native-executorch-supertonic

1---2license: openrail3language:4- en5- es6- fr7- pl8- el9- id10- zh11- ja12- ru13- tr14- da15- et16- hr17- pt18- vi19- sl20- cs21- it22- uk23- ko24- sv25- lv26- hu27- fi28- ar29- nl30- sk31- bg32- hi33- de34- lt35base_model:36- Supertone/supertonic-337pipeline_tag: text-to-speech38library_name: executorch39---40 41# supertonic42 43This repository hosts the **supertonic** models exported for the44[React Native ExecuTorch](https://www.npmjs.com/package/react-native-executorch)45library as ExecuTorch `.pte` programs, ready to run on device.46 47## Variants48 49| Path | Component | Backend | Precision |50| --- | --- | --- | --- |51| `mlx/duration_predictor_mlx_fp32.pte` | duration_predictor | mlx | fp32 |52| `mlx/text_encoder_mlx_fp32.pte` | text_encoder | mlx | fp32 |53| `mlx/vector_estimator_mlx_fp32.pte` | vector_estimator | mlx | fp32 |54| `mlx/vocoder_mlx_fp32.pte` | vocoder | mlx | fp32 |55| `vulkan/duration_predictor_vulkan_fp16.pte` | duration_predictor | vulkan | fp16 |56| `vulkan/text_encoder_vulkan_fp16.pte` | text_encoder | vulkan | fp16 |57| `vulkan/vector_estimator_vulkan_fp16.pte` | vector_estimator | vulkan | fp16 |58| `vulkan/vocoder_vulkan_fp16.pte` | vocoder | vulkan | fp16 |59| `xnnpack/duration_predictor_xnnpack_fp32.pte` | duration_predictor | xnnpack | fp32 |60| `xnnpack/text_encoder_xnnpack_fp32.pte` | text_encoder | xnnpack | fp32 |61| `xnnpack/vector_estimator_xnnpack_fp32.pte` | vector_estimator | xnnpack | fp32 |62| `xnnpack/vocoder_xnnpack_fp32.pte` | vocoder | xnnpack | fp32 |63 64## Unshipped backends65 66A backend directory carrying a `NOTES.md` and no `.pte` was exported and67not published. Its note records what went wrong.68 69- `coreml` ([why](https://huggingface.co/software-mansion/react-native-executorch-supertonic/blob/main/coreml/NOTES.md))70 71## Repository structure72 73```74config.json                                  32 B75coreml/NOTES.md                              879 B76mlx/config.json                              6.7 kB77mlx/duration_predictor_mlx_fp32.pte          3.4 MB78mlx/text_encoder_mlx_fp32.pte                34.4 MB79mlx/vector_estimator_mlx_fp32.pte            244 MB80mlx/vocoder_mlx_fp32.pte                     96.7 MB81unicode_indexer.json                         271 kB82voice_styles/F1.json                         285 kB83voice_styles/F2.json                         286 kB84voice_styles/F3.json                         284 kB85voice_styles/F4.json                         285 kB86voice_styles/F5.json                         285 kB87voice_styles/M1.json                         285 kB88voice_styles/M2.json                         285 kB89voice_styles/M3.json                         283 kB90voice_styles/M4.json                         285 kB91voice_styles/M5.json                         285 kB92voices/F1.bin                                50.5 kB93voices/F2.bin                                50.5 kB94voices/F3.bin                                50.5 kB95voices/F4.bin                                50.5 kB96voices/F5.bin                                50.5 kB97voices/M1.bin                                50.5 kB98voices/M2.bin                                50.5 kB99voices/M3.bin                                50.5 kB100voices/M4.bin                                50.5 kB101voices/M5.bin                                50.5 kB102vulkan/config.json                           4.1 kB103vulkan/duration_predictor_vulkan_fp16.pte    1.7 MB104vulkan/text_encoder_vulkan_fp16.pte          17.3 MB105vulkan/vector_estimator_vulkan_fp16.pte      122 MB106vulkan/vocoder_vulkan_fp16.pte               48.4 MB107xnnpack/config.json                          6.7 kB108xnnpack/duration_predictor_xnnpack_fp32.pte  3.4 MB109xnnpack/text_encoder_xnnpack_fp32.pte        34.5 MB110xnnpack/vector_estimator_xnnpack_fp32.pte    245 MB111xnnpack/vocoder_xnnpack_fp32.pte             96.7 MB112```113 114## Compatibility115 116These files are published for the **ExecuTorch v1.4.1** runtime. ExecuTorch117gives no forward compatibility guarantee, so an older runtime may fail to load118them.119 120To use them in React Native ExecuTorch, pass the model constant shipped in the121library's model registry to the corresponding task pipeline. See the122[documentation](https://docs.swmansion.com/react-native-executorch/docs/fundamentals/downloading-models).123 124To load these files in your own ExecuTorch runtime, read the125[compatibility note](https://github.com/pytorch/executorch/blob/main/runtime/COMPATIBILITY.md)126first.127