| { |
| "model_type": "magpie-tts", |
| "engine": "magpie", |
| "description": "NVIDIA Magpie-TTS Multilingual 357M exported to ONNX for phoonnx", |
| "source_model": "nvidia/magpie_tts_multilingual_357m", |
| "source_revision": "v2607", |
| "license": "nvidia-open-model-license", |
| "sample_rate": 22050, |
| "codec": "nvidia/nemo-nano-codec-22khz-1.89kbps-21.5fps", |
| "languages": ["ar", "de", "en", "es", "fr", "hi", "it", "ja", "ko", "pt", "vi", "zh"], |
| "speakers": {"Aria": 0, "Jason": 1, "John": 2, "Leo": 3, "Sofia": 4}, |
| "opset": 18, |
| "graphs": { |
| "text_encoder": {"file": "text_encoder.onnx", "inputs": ["text", "text_mask"], "outputs": ["encoder_out"]}, |
| "cross_kv": {"file": "cross_kv.onnx", "inputs": ["cond"], "outputs": ["cross_k", "cross_v"]}, |
| "decoder_step": {"file": "decoder_step.onnx", "inputs": ["x", "pos", "self_k", "self_v", "cross_k", "cross_v", "cond_mask", "attn_prior"], "outputs": ["logits", "dec_out", "new_self_k", "new_self_v", "cross_attn_probs"]}, |
| "local_step": {"file": "local_step.onnx", "inputs": ["h", "pos", "cache_k", "cache_v", "cb"], "outputs": ["logits", "new_k", "new_v"]}, |
| "audio_embed": {"file": "audio_embed.onnx", "inputs": ["codes"], "outputs": ["emb"]}, |
| "lt_embed": {"file": "lt_embed.onnx", "inputs": ["tok", "cb"], "outputs": ["emb"]}, |
| "codec_decoder": {"file": "codec_decoder.onnx", "inputs": ["codes"], "outputs": ["audio"]} |
| }, |
| "assets": { |
| "context_embeddings": "assets/context_embeddings.npy", |
| "tokenizer": "assets/tokenizer.json", |
| "speakers": "assets/speakers.json", |
| "config": "assets/magpie_config.json" |
| } |
| } |
|
|
