Instructions to use soniqo/Chatterbox-LiteRT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LiteRT
How to use soniqo/Chatterbox-LiteRT with LiteRT:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Chatterbox
How to use soniqo/Chatterbox-LiteRT with Chatterbox:
# pip install chatterbox-tts import torchaudio as ta from chatterbox.tts import ChatterboxTTS model = ChatterboxTTS.from_pretrained(device="cuda") text = "Ezreal and Jinx teamed up with Ahri, Yasuo, and Teemo to take down the enemy's Nexus in an epic late-game pentakill." wav = model.generate(text) ta.save("test-1.wav", wav, model.sr) # If you want to synthesize with a different voice, specify the audio prompt AUDIO_PROMPT_PATH="YOUR_FILE.wav" wav = model.generate(text, audio_prompt_path=AUDIO_PROMPT_PATH) ta.save("test-2.wav", wav, model.sr) - Notebooks
- Google Colab
- Kaggle
| { | |
| "architecture": "chatterbox", | |
| "model": "ResembleAI/chatterbox", | |
| "format": "litert", | |
| "variant": "fp16-multilingual", | |
| "deployment_complete": true, | |
| "voice": "default (zero-shot); cloning = follow-up (needs voice-encoder + s3-tokenizer + cond_enc graphs)", | |
| "sample_rate": 24000, | |
| "languages": [ | |
| "ar", | |
| "da", | |
| "de", | |
| "el", | |
| "en", | |
| "es", | |
| "fi", | |
| "fr", | |
| "he", | |
| "hi", | |
| "it", | |
| "ja", | |
| "ko", | |
| "ms", | |
| "nl", | |
| "no", | |
| "pl", | |
| "pt", | |
| "ru", | |
| "sv", | |
| "sw", | |
| "tr", | |
| "zh" | |
| ], | |
| "t3": { | |
| "hidden_size": 1024, | |
| "num_layers": 30, | |
| "num_kv_heads": 16, | |
| "head_dim": 64, | |
| "text_vocab": 2454, | |
| "speech_vocab": 8194, | |
| "start_speech_token": 6561, | |
| "stop_speech_token": 6562, | |
| "start_text_token": 255, | |
| "stop_text_token": 0, | |
| "cond_emb_len": 34, | |
| "prefill_text_pad": 478, | |
| "prefill_total": 512, | |
| "max_cache_tokens": 1536, | |
| "max_generated_tokens": 1000, | |
| "cfg_weight": 0.5, | |
| "temperature": 0.8, | |
| "batch": 1, | |
| "pipeline": "prefill(text_token_ids, cond_emb) -> cache; step(speech_token_id, speech_pos_idx, cache_position, cache) loop; BOS(6561) is step 0", | |
| "cache_layout": "per-layer key[1,16,cache_len,64] + value[1,16,cache_len,64]", | |
| "cfg_note": "batch=1; run cond + uncond (text_emb[1].zero_()) as two calls, combine: logits = cond + w*(cond-uncond)" | |
| }, | |
| "flow": { | |
| "encoder_max_tokens": 512, | |
| "estimator_max_frames": 1024, | |
| "inference_steps": 10, | |
| "solver": "euler", | |
| "cfg_rate": 0.7, | |
| "encoder_inputs": "token_ids (prompt_token ++ speech_tokens), token_len; folds input_embedding + conformer + encoder_proj", | |
| "token_mel_ratio": 2, | |
| "prompt_feat_frames": 314 | |
| }, | |
| "tokenizer": "grapheme_mtl_merged_expanded_v1.json", | |
| "cond_emb": { | |
| "file": "cond_emb.bin", | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 34, | |
| 1024 | |
| ] | |
| }, | |
| "files": { | |
| "t3_prefill": "chatterbox-t3-prefill.tflite", | |
| "t3_step": "chatterbox-t3-step.tflite", | |
| "flow_encoder": "chatterbox-flow-encoder.tflite", | |
| "flow_estimator": "chatterbox-flow-estimator.tflite", | |
| "hift": "chatterbox-hift.tflite" | |
| }, | |
| "graph_variants": { | |
| "t3_prefill": "fp16", | |
| "t3_step": "fp16", | |
| "flow_encoder": "fp32", | |
| "flow_estimator": "fp32", | |
| "hift": "fp32" | |
| }, | |
| "inputs": { | |
| "t3_prefill": { | |
| "serving_default_args_0:0": { | |
| "dtype": "int64", | |
| "shape": [ | |
| 1, | |
| 478 | |
| ] | |
| }, | |
| "serving_default_args_1:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 34, | |
| 1024 | |
| ] | |
| } | |
| }, | |
| "t3_step": { | |
| "serving_default_args_22:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_33:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_42:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_28:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_21:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_38:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_30:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_58:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_5:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_26:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_12:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_46:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_40:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_34:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_2:0": { | |
| "dtype": "int64", | |
| "shape": [ | |
| 1 | |
| ] | |
| }, | |
| "serving_default_args_25:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_62:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_60:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_31:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_37:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_53:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_3:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_52:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_24:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_54:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_29:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_19:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_9:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_0:0": { | |
| "dtype": "int64", | |
| "shape": [ | |
| 1, | |
| 1 | |
| ] | |
| }, | |
| "serving_default_args_14:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_43:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_49:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_56:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_50:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_10:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_45:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_39:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_36:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_16:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_20:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_11:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_6:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_48:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_4:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_51:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_13:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_18:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_41:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_8:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_15:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_47:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_59:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_35:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_44:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_1:0": { | |
| "dtype": "int64", | |
| "shape": [ | |
| 1, | |
| 1 | |
| ] | |
| }, | |
| "serving_default_args_27:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_32:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_55:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_17:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_61:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_7:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_23:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "serving_default_args_57:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| } | |
| }, | |
| "flow_encoder": { | |
| "serving_default_args_0:0": { | |
| "dtype": "int64", | |
| "shape": [ | |
| 1, | |
| 512 | |
| ] | |
| }, | |
| "serving_default_args_1:0": { | |
| "dtype": "int32", | |
| "shape": [ | |
| 1 | |
| ] | |
| } | |
| }, | |
| "flow_estimator": { | |
| "serving_default_args_5:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 80, | |
| 1024 | |
| ] | |
| }, | |
| "serving_default_args_2:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 80, | |
| 1024 | |
| ] | |
| }, | |
| "serving_default_args_3:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2 | |
| ] | |
| }, | |
| "serving_default_args_0:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 80, | |
| 1024 | |
| ] | |
| }, | |
| "serving_default_args_4:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 80 | |
| ] | |
| }, | |
| "serving_default_args_1:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 1, | |
| 1024 | |
| ] | |
| } | |
| }, | |
| "hift": { | |
| "serving_default_args_0:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 80, | |
| 1024 | |
| ] | |
| } | |
| } | |
| }, | |
| "outputs": { | |
| "t3_prefill": { | |
| "StatefulPartitionedCall:33": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:60": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:34": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:42": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:1": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:22": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:40": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:44": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:21": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:49": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:30": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:16": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:46": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:10": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:29": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:13": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:37": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:31": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:58": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:48": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:32": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:6": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:43": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:54": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:4": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:28": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:38": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:45": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:27": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:57": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:52": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:39": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:23": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:55": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:51": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:15": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:35": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:14": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:20": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:25": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:41": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:5": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:18": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:19": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:9": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:11": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 512, | |
| 8194 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:17": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:12": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:7": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:59": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:47": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:3": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:2": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:36": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:50": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:8": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:26": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:53": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:56": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:24": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 512, | |
| 64 | |
| ] | |
| } | |
| }, | |
| "t3_step": { | |
| "StatefulPartitionedCall:33": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:60": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:34": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:42": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:1": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:22": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:40": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:44": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:21": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:49": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:30": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:16": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:46": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:10": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:29": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:13": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:37": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:31": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:58": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:48": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:32": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:6": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:43": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:54": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:4": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:28": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:38": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:45": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:27": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:57": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:52": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:39": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:23": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:55": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:51": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:15": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:35": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:14": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:20": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:25": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:41": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:5": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:18": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:19": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:9": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:11": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 1, | |
| 8194 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:17": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:12": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:7": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:59": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:47": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:3": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:2": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:36": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:50": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:8": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:26": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:53": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:56": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:24": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 16, | |
| 1536, | |
| 64 | |
| ] | |
| } | |
| }, | |
| "flow_encoder": { | |
| "StatefulPartitionedCall:1": { | |
| "dtype": "bool", | |
| "shape": [ | |
| 1, | |
| 1, | |
| 1024 | |
| ] | |
| }, | |
| "StatefulPartitionedCall:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 80, | |
| 1024 | |
| ] | |
| } | |
| }, | |
| "flow_estimator": { | |
| "StatefulPartitionedCall:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 2, | |
| 80, | |
| 1024 | |
| ] | |
| } | |
| }, | |
| "hift": { | |
| "StatefulPartitionedCall:0": { | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 491520 | |
| ] | |
| } | |
| } | |
| }, | |
| "export": { | |
| "source_repo": "ResembleAI/chatterbox", | |
| "toolchain": "ai_edge_torch 0.6 + ai_edge_quantizer (dynamic_legacy_wi8_afp32 on T3)", | |
| "embeddings": "folded into T3 graphs (token-ID inputs); text/speech emb + learned pos inside graph", | |
| "kv_cache": "graph IO, functional StaticLayer.update; prefill cache=total, copy first prefix_len into step cache 1536", | |
| "stft": "iSTFT/STFT re-expressed as fixed-basis Conv1d/ConvTranspose1d", | |
| "flow_embeddings": "folded into flow_encoder_v2 (token-ID); input_embedding + encoder_proj inside graph", | |
| "t3_precision": "fp16 (float_casting); int8 too lossy for Arabic" | |
| }, | |
| "flow_cond": { | |
| "voice": "default (zero-shot)", | |
| "spks": { | |
| "file": "spks.bin", | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 80 | |
| ] | |
| }, | |
| "prompt_token": { | |
| "file": "prompt_token.bin", | |
| "dtype": "int32", | |
| "shape": [ | |
| 1, | |
| 157 | |
| ] | |
| }, | |
| "prompt_feat": { | |
| "file": "prompt_feat.bin", | |
| "dtype": "float32", | |
| "shape": [ | |
| 1, | |
| 314, | |
| 80 | |
| ] | |
| }, | |
| "note": "host: full_token = prompt_token ++ speech_tokens; flow_encoder -> mu; 10x Euler CFM (estimator batch-2 CFG, dxdt=1.7*cond-0.7*uncond, t_span=1-cos(linspace(0,1,11)*pi/2)) with x init = randn; mel = x[:,:,prompt_feat_frames:]; hift -> 24k pcm" | |
| }, | |
| "precision_note": "T3 graphs are fp16 (multilingual, incl. Arabic). int8 T3 is ~2x smaller but English-quality only (diverges on Arabic); available on request." | |
| } |