Home / FunAudioLLM / CosyVoice
Repository Analysis

FunAudioLLM/CosyVoice

Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability.

15.7 Moderate AI signal View on GitHub
15.7
Adjusted Score
15.7
Raw Score
100%
Time Factor
2026-05-25
Last Push
21,333
Stars
Python
Language
26,042
Lines of Code
136
Files
259
Pattern Hits
2026-05-31
Scan Date

Score History

Severity Breakdown

CRITICAL 0HIGH 33MEDIUM 14LOW 212

Pattern Findings

259 matches across 11 categories. Click a row to expand file-level details.

Cross-File Repetition21 hits · 105 pts
SeverityFileLineSnippet
HIGHcosyvoice/transformer/subsampling.py0input x. args: x (torch.tensor): input tensor (#batch, time, idim). x_mask (torch.tensor): input mask (#batch, 1, time).
HIGHcosyvoice/transformer/subsampling.py0input x. args: x (torch.tensor): input tensor (#batch, time, idim). x_mask (torch.tensor): input mask (#batch, 1, time).
HIGHcosyvoice/transformer/subsampling.py0input x. args: x (torch.tensor): input tensor (#batch, time, idim). x_mask (torch.tensor): input mask (#batch, 1, time).
HIGHruntime/triton_trtllm/token2wav_dit.py0example usage cuda_visible_devices=0 \ python3 token2wav.py --enable-trt || exit 1
HIGHruntime/triton_trtllm/token2wav.py0example usage cuda_visible_devices=0 \ python3 token2wav.py --enable-trt || exit 1
HIGH…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py0example usage cuda_visible_devices=0 \ python3 token2wav.py --enable-trt || exit 1
HIGHruntime/triton_trtllm/offline_inference.py0extract speech ids from token strings like <|s_23456|>
HIGHruntime/triton_trtllm/infer_cosyvoice3.py0extract speech ids from token strings like <|s_23456|>
HIGHexamples/grpo/cosyvoice2/infer_dataset.py0extract speech ids from token strings like <|s_23456|>
HIGHruntime/triton_trtllm/model_repo/cosyvoice2/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGHruntime/triton_trtllm/model_repo/token2wav/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…me/triton_trtllm/model_repo/audio_tokenizer/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…time/triton_trtllm/model_repo/token2wav_dit/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…/triton_trtllm/model_repo/speaker_embedding/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…rtllm/model_repo_cosyvoice3/audio_tokenizer/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…llm/model_repo_cosyvoice3/speaker_embedding/1/model.py0initialize the model. args: args: dictionary containing model configuration
HIGH…me/triton_trtllm/model_repo/audio_tokenizer/1/model.py0triton python model for audio tokenization. this model takes reference audio input and extracts semantic tokens using s3
HIGH…/triton_trtllm/model_repo/speaker_embedding/1/model.py0triton python model for audio tokenization. this model takes reference audio input and extracts semantic tokens using s3
HIGH…rtllm/model_repo_cosyvoice3/audio_tokenizer/1/model.py0triton python model for audio tokenization. this model takes reference audio input and extracts semantic tokens using s3
HIGH…llm/model_repo_cosyvoice3/speaker_embedding/1/model.py0triton python model for audio tokenization. this model takes reference audio input and extracts semantic tokens using s3
Over-Commented Block70 hits · 70 pts
SeverityFileLineSnippet
LOWwebui.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Liu Yue)
LOWcosyvoice/llm/llm.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWcosyvoice/bin/export_jit.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/bin/average_model.py1# Copyright (c) 2020 Mobvoi Inc (Di Wu)
LOWcosyvoice/bin/train.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/bin/export_onnx.py1# Copyright (c) 2024 Antgroup Inc (authors: Zhoubofan, hexisyztem@icloud.com)
LOWcosyvoice/dataset/processor.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/dataset/dataset.py1# Copyright (c) 2021 Mobvoi Inc. (authors: Binbin Zhang)
LOWcosyvoice/hifigan/generator.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Kai Hu)
LOWcosyvoice/hifigan/f0_predictor.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Kai Hu)
LOWcosyvoice/transformer/decoder.py1# Copyright (c) 2021 Mobvoi Inc. (authors: Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/attention.py1# Copyright (c) 2019 Shigeki Karita
LOWcosyvoice/transformer/attention.py161 Returns:
LOWcosyvoice/transformer/attention.py281 # cache(1, head, 0, d_k * 2) (16/-1, -1/-1, 16/0 mode)
LOWcosyvoice/transformer/subsampling.py1# Copyright (c) 2021 Mobvoi Inc (Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/convolution.py1# Copyright (c) 2020 Mobvoi Inc. (authors: Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/upsample_encoder.py1# Copyright (c) 2021 Mobvoi Inc (Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/embedding.py1# Copyright (c) 2020 Mobvoi Inc. (authors: Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/encoder_layer.py1# Copyright (c) 2021 Mobvoi Inc (Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/encoder.py1# Copyright (c) 2021 Mobvoi Inc (Binbin Zhang, Di Wu)
LOWcosyvoice/transformer/activation.py1# Copyright (c) 2020 Johns Hopkins University (Shinji Watanabe)
LOWcosyvoice/transformer/positionwise_feed_forward.py1# Copyright (c) 2019 Shigeki Karita
LOWcosyvoice/transformer/decoder_layer.py1# Copyright (c) 2019 Shigeki Karita
LOWcosyvoice/transformer/label_smoothing_loss.py1# Copyright (c) 2019 Shigeki Karita
LOWcosyvoice/vllm/cosyvoice2.py1# SPDX-License-Identifier: Apache-2.0
LOWcosyvoice/utils/class_utils.py1# Copyright [2023-11-28] <sxc19@mails.tsinghua.edu.cn, Xingchen Song>
LOWcosyvoice/utils/mask.py1# Copyright (c) 2019 Shigeki Karita
LOWcosyvoice/utils/train_utils.py1# Copyright (c) 2021 Mobvoi Inc. (authors: Binbin Zhang)
LOWcosyvoice/utils/common.py1# Copyright (c) 2020 Mobvoi Inc (Binbin Zhang)
LOWcosyvoice/utils/file_utils.py1# Copyright (c) 2021 Mobvoi Inc. (authors: Binbin Zhang)
LOWcosyvoice/utils/scheduler.py1# Copyright (c) 2020 Mobvoi Inc (Binbin Zhang)
LOWcosyvoice/utils/frontend_utils.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWcosyvoice/utils/executor.py1# Copyright (c) 2020 Mobvoi Inc (Binbin Zhang)
LOWcosyvoice/cli/frontend.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/cli/cosyvoice.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/cli/model.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWcosyvoice/flow/decoder.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWcosyvoice/flow/flow.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWcosyvoice/flow/length_regulator.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWcosyvoice/flow/flow_matching.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu, Zhihao Du)
LOWtools/make_parquet_list.py1#!/usr/bin/env python3
LOWtools/extract_speech_token.py1#!/usr/bin/env python3
LOWtools/extract_embedding.py1#!/usr/bin/env python3
LOWruntime/python/grpc/server.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWruntime/python/grpc/client.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWruntime/python/fastapi/server.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWruntime/python/fastapi/client.py1# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
LOWruntime/triton_trtllm/token2wav_dit.py1# SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
LOWruntime/triton_trtllm/run_stepaudio2_dit_token2wav.sh101if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
LOWruntime/triton_trtllm/client_http.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOWruntime/triton_trtllm/offline_inference.py1# SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
LOWruntime/triton_trtllm/client_grpc.py1# Copyright 2022 Xiaomi Corp. (authors: Fangjun Kuang)
LOWruntime/triton_trtllm/token2wav.py1# SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
LOWruntime/triton_trtllm/model_repo/cosyvoice2/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOWruntime/triton_trtllm/model_repo/token2wav/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOW…me/triton_trtllm/model_repo/audio_tokenizer/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOW…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py1# SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
LOW…/triton_trtllm/model_repo/speaker_embedding/1/model.py1# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
10 more matches not shown…
Unused Imports57 hits · 57 pts
SeverityFileLineSnippet
LOWwebui.py19
LOWwebui.py22
LOWcosyvoice/llm/llm.py18
LOWcosyvoice/llm/llm.py20
LOWcosyvoice/bin/export_jit.py15
LOWcosyvoice/bin/train.py15
LOWcosyvoice/bin/export_onnx.py16
LOWcosyvoice/vllm/cosyvoice2.py35
LOWcosyvoice/flow/DiT/modules.py11
LOWcosyvoice/flow/DiT/dit.py11
LOWruntime/triton_trtllm/offline_inference.py34
LOWruntime/triton_trtllm/offline_inference.py39
LOWruntime/triton_trtllm/streaming_inference.py7
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py12
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py14
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py19
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py20
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py28
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py29
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py32
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py32
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py32
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py42
LOWruntime/triton_trtllm/model_repo/token2wav/1/model.py40
LOW…me/triton_trtllm/model_repo/audio_tokenizer/1/model.py33
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py28
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py31
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py31
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py35
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py39
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py40
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py41
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py41
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py42
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py43
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py44
LOW…/triton_trtllm/model_repo/speaker_embedding/1/model.py33
LOW…iton_trtllm/model_repo_cosyvoice3/token2wav/1/model.py7
LOW…rtllm/model_repo_cosyvoice3/audio_tokenizer/1/model.py33
LOW…llm/model_repo_cosyvoice3/speaker_embedding/1/model.py33
LOWruntime/triton_trtllm/scripts/test_llm.py19
LOW…time/triton_trtllm/scripts/convert_cosyvoice3_to_hf.py39
LOW…time/triton_trtllm/scripts/convert_cosyvoice3_to_hf.py39
LOW…time/triton_trtllm/scripts/convert_cosyvoice3_to_hf.py75
LOWexamples/grpo/cosyvoice2/prepare_data.py20
LOWexamples/grpo/cosyvoice2/infer_dataset.py30
LOWexamples/grpo/cosyvoice2/infer_dataset.py33
LOWexamples/grpo/cosyvoice2/infer_dataset.py37
LOWexamples/grpo/cosyvoice2/infer_dataset.py43
LOWexamples/grpo/cosyvoice2/pretrained_to_huggingface.py27
LOWexamples/grpo/cosyvoice2/token2wav_asr_server.py20
LOWexamples/grpo/cosyvoice2/token2wav_asr_server.py25
LOWexamples/grpo/cosyvoice2/token2wav_asr_server.py27
LOWexamples/grpo/cosyvoice2/token2wav_asr_server.py33
LOWexamples/grpo/cosyvoice2/reward_tts.py19
LOWexamples/grpo/cosyvoice2/reward_tts.py119
LOW…amples/grpo/cosyvoice2/scripts/offline-decode-files.py87
Excessive Try-Catch Wrapping22 hits · 40 pts
SeverityFileLineSnippet
LOWcosyvoice/dataset/processor.py51 except Exception as ex:
MEDIUMruntime/triton_trtllm/client_grpc.py100 print(f"An error occurred in the stream callback: {error}")
MEDIUMruntime/triton_trtllm/client_grpc.py576 print(f"Error: Audio file not found for item {i}: {item['audio_filepath']}")
LOWruntime/triton_trtllm/client_grpc.py577 except Exception as e:
MEDIUMruntime/triton_trtllm/client_grpc.py578 print(f"Error processing item {i} ({item['target_audio_path']}): {e}")
LOWruntime/triton_trtllm/client_grpc.py588 except Exception as e:
LOWruntime/triton_trtllm/client_grpc.py742 except Exception as e:
LOWruntime/triton_trtllm/client_grpc.py903 except Exception as e:
LOWruntime/triton_trtllm/client_grpc.py910 except Exception as e:
MEDIUMruntime/triton_trtllm/client_grpc.py911 print(f"Error closing async stats client: {e}")
LOWruntime/triton_trtllm/client_grpc.py918 except Exception as e:
MEDIUMruntime/triton_trtllm/client_grpc.py919 print(f"An error occurred in main: {e}")
MEDIUMruntime/triton_trtllm/client_grpc.py915def run_main():
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py427 except Exception as e:
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py483 except Exception as e:
LOWruntime/triton_trtllm/scripts/convert_checkpoint.py293 except Exception as e:
LOWexamples/grpo/cosyvoice2/reward_tts.py106 except Exception as e:
LOWexamples/grpo/cosyvoice2/reward_tts.py212 except Exception as e:
MEDIUMexamples/grpo/cosyvoice2/reward_tts.py213 print(f"Error computing score: {e}")
MEDIUMexamples/grpo/cosyvoice2/reward_tts.py229 print(f"Error: File not found - {args.input}")
LOWexamples/grpo/cosyvoice2/reward_tts.py232 except Exception as e:
MEDIUMexamples/grpo/cosyvoice2/reward_tts.py233 print(f"Error: {e}")
Docstring Block Structure8 hits · 40 pts
SeverityFileLineSnippet
HIGHcosyvoice/utils/mask.py18 def subsequent_mask( size: int, device: torch.device = torch.device("cpu"), ) -> torch.Tensor: """C
HIGHcosyvoice/utils/mask.py57Create mask for subsequent steps (size, size). This mask is used only in decoder which works in an auto-regressive
HIGHcosyvoice/utils/mask.py95Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: s
HIGHcosyvoice/utils/mask.py133Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: s
HIGHcosyvoice/utils/mask.py240Make mask tensor containing indices of padded part. See description of make_non_pad_mask. Args: length
HIGHcosyvoice/utils/common.py57Perform padding for the list of tensors. Args: xs (List): List of Tensors [(T_1, `*`), (T_2, `*`), ..., (T_
HIGHcosyvoice/flow/decoder.py211Forward pass of the UNet1DConditional model. Args: x (torch.Tensor): shape (batch_size, in_channels
HIGHcosyvoice/flow/decoder.py406Forward pass of the UNet1DConditional model. Args: x (torch.Tensor): shape (batch_size, in_channels
Deep Nesting37 hits · 37 pts
SeverityFileLineSnippet
LOWwebui.py50
LOWcosyvoice/llm/llm.py302
LOWcosyvoice/llm/llm.py505
LOWcosyvoice/llm/llm.py552
LOWcosyvoice/bin/average_model.py43
LOWcosyvoice/dataset/processor.py31
LOWcosyvoice/utils/mask.py161
LOWcosyvoice/utils/train_utils.py111
LOWcosyvoice/utils/train_utils.py291
LOWcosyvoice/cli/model.py101
LOWcosyvoice/cli/model.py175
LOWcosyvoice/flow/decoder.py196
LOWruntime/python/grpc/client.py30
LOWruntime/triton_trtllm/token2wav_dit.py175
LOWruntime/triton_trtllm/offline_inference.py325
LOWruntime/triton_trtllm/client_grpc.py112
LOWruntime/triton_trtllm/client_grpc.py159
LOWruntime/triton_trtllm/client_grpc.py412
LOWruntime/triton_trtllm/client_grpc.py501
LOWruntime/triton_trtllm/client_grpc.py687
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py193
LOWruntime/triton_trtllm/token2wav.py127
LOWruntime/triton_trtllm/infer_cosyvoice3.py219
LOWruntime/triton_trtllm/model_repo/cosyvoice2/1/model.py315
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py121
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py295
LOW…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py175
LOW…/triton_trtllm/model_repo/speaker_embedding/1/model.py91
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py84
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py305
LOW…llm/model_repo_cosyvoice3/speaker_embedding/1/model.py91
LOWruntime/triton_trtllm/scripts/fill_template.py6
LOWruntime/triton_trtllm/scripts/convert_checkpoint.py152
LOWruntime/triton_trtllm/scripts/convert_checkpoint.py186
LOWruntime/triton_trtllm/scripts/convert_checkpoint.py282
LOWexamples/grpo/cosyvoice2/infer_dataset.py282
LOW…amples/grpo/cosyvoice2/scripts/offline-decode-files.py137
Hyper-Verbose Identifiers34 hits · 34 pts
SeverityFileLineSnippet
LOWcosyvoice/transformer/decoder.py178 def forward_layers_checkpointed(self, x: torch.Tensor,
LOWcosyvoice/transformer/encoder.py173 def forward_layers_checkpointed(self, xs: torch.Tensor,
LOWcosyvoice/utils/mask.py89def subsequent_chunk_mask_deprecated(
LOWcosyvoice/utils/train_utils.py53def init_dataset_and_dataloader(args, configs, gan, dpo):
LOWcosyvoice/utils/train_utils.py72def check_modify_and_save_config(args, configs):
LOWcosyvoice/utils/train_utils.py111def init_optimizer_and_scheduler(args, configs, model, gan):
LOWcosyvoice/utils/scheduler.py397def _linear_warmup_with_cosine_annealing(max_lr, warmup_steps, step,
LOWcosyvoice/utils/scheduler.py543 def _get_linear_warmup_with_cosine_annealing_lr(self, step):
LOWcosyvoice/cli/frontend.py89 def _extract_text_token_generator(self, text_generator):
LOWruntime/triton_trtllm/token2wav_dit.py237 def get_trt_kwargs_dynamic_batch(self, opt_batch_size=2, max_batch_size=64, streaming=False):
LOWruntime/triton_trtllm/token2wav_dit.py264 def prompt_audio_tokenization(self, prompt_audios_list: list[torch.Tensor]) -> list[list[int]]:
LOWruntime/triton_trtllm/token2wav_dit.py371 def get_prompt_audio_cache_for_streaming_tts(
LOWruntime/triton_trtllm/offline_inference.py63async def send_batch_requests_async(api_base, model_name, chats, temperature, top_p, top_k):
LOWruntime/triton_trtllm/client_grpc.py359def prepare_request_input_output(
LOWruntime/triton_trtllm/client_grpc.py412def run_sync_streaming_inference(
LOWruntime/triton_trtllm/token2wav_cosyvoice3.py219 def prompt_audio_tokenization(self, prompt_audios_list):
LOWruntime/triton_trtllm/token2wav.py185 def get_trt_kwargs_dynamic_batch(self, opt_bs=2, max_batch_size=64):
LOWruntime/triton_trtllm/token2wav.py193 def prompt_audio_tokenization(self, prompt_audios_list: list[torch.Tensor]) -> list[list[int]]:
LOWruntime/triton_trtllm/infer_cosyvoice3.py45async def send_batch_requests_async(api_base, model_name, chats, temperature, top_p, top_k):
LOWruntime/triton_trtllm/model_repo/cosyvoice2/1/model.py201 def forward_speaker_embedding(self, wav):
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py52def parse_speech_token_string(response_text: str) -> List[int]:
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py108 def _convert_speech_tokens_to_str(self, speech_tokens: Union[torch.Tensor, List]) -> str:
LOW…ime/triton_trtllm/model_repo/cosyvoice2_dit/1/model.py207 def forward_speaker_embedding(self, wav):
LOW…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py237 def get_trt_kwargs_dynamic_batch(self, opt_batch_size=2, max_batch_size=64, streaming=False):
LOW…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py264 def prompt_audio_tokenization(self, prompt_audios_list: list[torch.Tensor]) -> list[list[int]]:
LOW…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py371 def get_prompt_audio_cache_for_streaming_tts(
LOW…time/triton_trtllm/model_repo/token2wav_dit/1/model.py56def get_spk_id_from_prompt_audio(tensor: torch.Tensor) -> str:
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py25def parse_speech_token_string(response_text):
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py72 def _convert_speech_tokens_to_str(self, speech_tokens):
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py181 def forward_speaker_embedding(self, wav):
LOW…ton_trtllm/model_repo_cosyvoice3/cosyvoice3/1/model.py305 async def _process_request_streaming(self, request):
LOWruntime/triton_trtllm/scripts/convert_checkpoint.py186def update_quant_config_from_hf(quant_config, hf_config,
LOWexamples/grpo/cosyvoice2/token2wav_asr_server.py116def get_random_prompt_from_dataset(dataset):
LOW…amples/grpo/cosyvoice2/scripts/offline-decode-files.py588def normalize_text_alimeeting(text: str) -> str:
Cross-Language Confusion4 hits · 10 pts
SeverityFileLineSnippet
HIGHruntime/triton_trtllm/token2wav_dit.py17 python3 token2wav.py --enable-trt || exit 1
HIGHruntime/triton_trtllm/token2wav_cosyvoice3.py3 python3 token2wav_cosyvoice3.py --enable-trt || exit 1
HIGHruntime/triton_trtllm/token2wav.py17 python3 token2wav.py --enable-trt || exit 1
HIGH…ton_trtllm/model_repo/token2wav_dit/1/token2wav_dit.py17 python3 token2wav.py --enable-trt || exit 1
AI Slop Vocabulary3 hits · 8 pts
SeverityFileLineSnippet
LOWcosyvoice/transformer/convolution.py125 # It's better we just return None if no cache is required,
MEDIUMcosyvoice/utils/executor.py45 # A context manager to be used in conjunction with an instance of
MEDIUMcosyvoice/utils/executor.py97 # A context manager to be used in conjunction with an instance of
Self-Referential Comments2 hits · 6 pts
SeverityFileLineSnippet
MEDIUM…time/triton_trtllm/model_repo/token2wav_dit/1/model.py64 # Create a SHA-256 hash of the byte string
MEDIUMexamples/grpo/cosyvoice2/run.sh9 # This function is from espnet
Slop Phrases1 hit · 3 pts
SeverityFileLineSnippet
MEDIUMexample.py60 # bistream usage, you can use generator as input, this is useful when using text llm model as input