chore: add config.json — model metadata + enable HF download counter

Without one of HF's default query files at the repo root (config.json,
config.yaml, hyperparams.yaml, params.json, meta.yaml), the Hub doesn't
register any downloads — HfApi reported 'downloads: 0' for this repo
because Pipeline.from_pretrained() pulls weights/*.safetensors but
never touches a recognised query file.

Adding config.json fixes the counter AND provides a single discoverable
metadata file:
- model_type, library_name, base_model, pipeline_tag
- the 4 sub-architectures (DP / TE / VE / vocoder)
- 31 supported languages (ISO codes)
- 13 voices (10 presets + 3 custom blends)
- inference config (5 Euler steps, CFG 4x cond - 3x uncond, default seed 99)
- measured RTF on M4 and M3 Ultra
- license trail (OpenRAIL-M weights + Apache-2.0 code)

Ref: https://huggingface.co/docs/hub/models-download-stats
This commit is contained in:
ambassadia
2026-05-20 16:13:40 +02:00
parent 052b24d0ac
commit ea1f5f2f01

58
config.json Normal file
View File

@@ -0,0 +1,58 @@
{
"model_type": "supertonic-3",
"library_name": "supertonic-3-mlx",
"base_model": "Supertone/supertonic-3",
"framework": "mlx",
"pipeline_tag": "text-to-speech",
"architectures": [
"DurationPredictor",
"TextEncoder",
"VectorEstimator",
"Vocoder"
],
"sample_rate": 44100,
"num_languages": 31,
"supported_languages": [
"en", "fr", "de", "es", "it", "pt", "ja", "ko", "zh", "ru",
"pl", "nl", "tr", "ar", "hi", "vi", "th", "id", "cs", "ro",
"hu", "el", "da", "sv", "fi", "no", "he", "uk", "bg", "hr", "sk"
],
"voices": {
"presets": ["F1", "F2", "F3", "F4", "F5", "M1", "M2", "M3", "M4", "M5"],
"custom": ["voix_sombre", "homme_moyen", "homme_clair"],
"total": 13
},
"inference": {
"euler_steps": 5,
"cfg_cond_scale": 4.0,
"cfg_uncond_scale": 3.0,
"default_seed": 99,
"supports_streaming": true,
"supports_voice_mixing": true
},
"performance_m4": {
"short_utterance_ms": 30,
"long_utterance_ms": 38,
"rtf_short": 76,
"rtf_long": 138,
"vs_onnx_sdk": "17-25x",
"vs_coreml": "2-3x"
},
"performance_m3_ultra": {
"rtf_short": 147,
"rtf_long": 185
},
"license": "openrail",
"license_link": "LICENSE",
"license_code": "Apache-2.0",
"license_code_link": "LICENSE-CODE",
"upstream_attribution": "Copyright (c) 2026 Supertone Inc."
}