| { | |
| "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json", | |
| "model": "kokoro_std_xnnpack_fp32", | |
| "family": "kokoro", | |
| "capabilities": [ | |
| "text-to-speech" | |
| ], | |
| "backend": "xnnpack", | |
| "license": "apache-2.0", | |
| "variants": [ | |
| { | |
| "file": null, | |
| "components": { | |
| "duration_predictor": "duration_predictor_std_xnnpack_fp32.pte", | |
| "synthesizer": "synthesizer_std_xnnpack_fp32.pte" | |
| }, | |
| "precision": "fp32", | |
| "methods": { | |
| "duration_predictor": { | |
| "forward_32": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 32 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 32 | |
| ], | |
| "dtype": "bool" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 32, | |
| 640 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| }, | |
| "forward_64": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 64 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 64 | |
| ], | |
| "dtype": "bool" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 64, | |
| 640 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| }, | |
| "forward_128": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 128 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 128 | |
| ], | |
| "dtype": "bool" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 128 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 128, | |
| 640 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| }, | |
| "synthesizer": { | |
| "forward": { | |
| "inputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "bool" | |
| }, | |
| { | |
| "shape": [ | |
| -1 | |
| ], | |
| "dtype": "int64" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| -1, | |
| 640 | |
| ], | |
| "dtype": "float32" | |
| }, | |
| { | |
| "shape": [ | |
| 1, | |
| 256 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "shape": [ | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "dtype": "float32" | |
| } | |
| ] | |
| } | |
| } | |
| } | |
| } | |
| ] | |
| } | |