| { | |
| "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json", | |
| "model": "supertonic3_vulkan_fp16", | |
| "family": "supertonic", | |
| "capabilities": [ | |
| "text-to-speech" | |
| ], | |
| "backend": "vulkan", | |
| "license": "openrail", | |
| "variants": [ | |
| { | |
| "file": null, | |
| "components": { | |
| "duration_predictor": "duration_predictor_vulkan_fp16.pte", | |
| "text_encoder": "text_encoder_vulkan_fp16.pte", | |
| "vector_estimator": "vector_estimator_vulkan_fp16.pte", | |
| "vocoder": "vocoder_vulkan_fp16.pte" | |
| }, | |
| "precision": "fp16", | |
| "methods": { | |
| "duration_predictor": { | |
| "forward": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 8, | |
| 16 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| }, | |
| "text_encoder": { | |
| "forward": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 50, | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 256, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| }, | |
| "vector_estimator": { | |
| "forward": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 144, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 256, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 50, | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 144, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| }, | |
| "vocoder": { | |
| "forward": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 144, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| } | |
| } | |
| } | |
| ] | |
| } | |