SmollerTTS / model.json
dunli5107's picture
Release SmollerTTS FP32 weights, browser runtime and model card
a4b7e51 verified
Raw History Blame Contribute Delete
1.55 kB
{
"architecture": "english-v9-acoustic-hifigan-v2",
"step": 10000,
"parameters": 2808792,
"acoustic_parameters": 1882807,
"vocoder_parameters": 925985,
"sample_rate": 22050,
"hop": 256,
"graphs": {
"encoder": {
"bytes": 5002113,
"sha256": "0872d67a5ebb0e3fb64e225f0c551c7f41dae65d5b97c1e7814da21163e40a22"
},
"frames": {
"bytes": 2592343,
"sha256": "7177cd29420ca1a65aa30bc79f33cd11f5ca26b96a6d0052fc80b961bc01784b"
},
"vocoder": {
"bytes": 3743259,
"sha256": "fad5accf376f23db956395fee1523f9cf1ffd99d53b5b0dbc4666d3f1b9ae743"
}
},
"checkpoint_sha256": "46d728f6b8cc8721d7c41607f63780016ed8c04d694dbcf9cc9863709c07399e",
"arithmetic": "Unfused PyTorch CPU reference (MKLDNN disabled); ONNX Runtime graph optimizations disabled in Python and WASM",
"quantization": "FP32; experimental INT8 variant rejected",
"name": "SmollerTTS",
"release": "fp32-v1",
"language": "en",
"speakers": 1,
"acoustic_initialization": "Random initialization; trained on LJSpeech",
"vocoder": {
"repository": "https://github.com/jik876/hifi-gan",
"variant": "LJ_V2",
"pretrained": true,
"generator_sha256": "3fac378c5918fb2c102733f21eeaa8e9a4ca6cda24dbfddc55bbb947c78d562f"
},
"pronunciation_frontend": {
"revision": "v27",
"automatic_splitting_default": true,
"punctuation_default": true,
"bytes": 2065502,
"sha256": "923e307f5ab200bd178ffdae44f5f49699667130952dcf68f8bdb763a62ed0b3"
}
}