MCPcopy Create free account

hub / github.com/Soul-AILab/SoulX-Podcast / functions

Functions193 in github.com/Soul-AILab/SoulX-Podcast

↓ 29 callersFunctioni18n
(key)
webui.py:200
↓ 14 callersMethod__init__
(self, dim, dim_out, groups=8)
soulxpodcast/models/modules/flow_components/estimator.py:350
↓ 11 callersFunctionadd_optional_chunk_mask
Apply optional mask for encoder. Args: xs (torch.Tensor): padded input, (B, L, D), L for max length mask (torch.Tensor): mask fo
soulxpodcast/models/modules/flow_components/upsample_encoder.py:43
↓ 8 callersMethod__init__
(self, channels: int, out_channels: int, stride: int = 2)
soulxpodcast/models/modules/flow_components/upsample_encoder.py:314
↓ 6 callersFunctionmask_to_bias
(mask: torch.Tensor, dtype: torch.dtype)
soulxpodcast/models/modules/flow_components/estimator.py:18
↓ 5 callersMethod__init__
(self, in_features, alpha=1.0, alpha_trainable=True, alpha_logscale=False)
soulxpodcast/models/modules/hifigan_components/layers.py:61
↓ 4 callersFunctioncheck_monologue_text
(text: str, prefix: str = None)
webui.py:204
↓ 4 callersFunctioncheck_monologue_text
(text: str, prefix: str = None)
soulxpodcast/utils/text.py:46
↓ 4 callersMethodforward_longform
( self, prompt_mels_for_llm, prompt_mels_lens_for_llm: torch.Tensor, prompt_text_token
soulxpodcast/models/soulxpodcast.py:50
↓ 4 callersFunctionget_service
获取全局服务实例
api/service.py:288
↓ 4 callersFunctionget_task_manager
获取全局任务管理器实例
api/tasks.py:215
↓ 4 callersMethodis_loaded
检查模型是否已加载
api/service.py:84
↓ 4 callersFunctionmake_pad_mask
Make mask tensor containing indices of padded part. See description of make_non_pad_mask. Args: lengths (torch.Tensor): Batch of len
soulxpodcast/models/modules/flow_components/upsample_encoder.py:121
↓ 3 callersMethodcreate_task
创建并加入队列
api/tasks.py:153
↓ 3 callersMethoddecode
(self, x: torch.Tensor, s: torch.Tensor = torch.zeros(1, 1, 0))
soulxpodcast/models/modules/hifigan.py:203
↓ 3 callersMethodfrom_initial_and_json
Create an instance from initial values and JSON data. Args: initial_values: Dictionary of initial values (highes
soulxpodcast/config.py:45
↓ 3 callersMethodgenerate
( self, prompt: list[str], sampling_param: SamplingParams, past_key_values=Non
soulxpodcast/engine/llm_engine.py:97
↓ 3 callersFunctionnormalize_text
(current_text)
soulxpodcast/utils/text.py:17
↓ 3 callersFunctionpodcast_format_parser
Parse the original multi-speaker podcast JSON to the target flattened format. The key will be a time-based unique ID. Args: data
soulxpodcast/utils/parser.py:20
↓ 3 callersMethodupdate_datasource
(self, data_list)
soulxpodcast/utils/dataloader.py:197
↓ 2 callersMethod_f02uv
(self, f0)
soulxpodcast/models/modules/hifigan_components/layers.py:300
↓ 2 callersFunctionaudio_volume_normalize
Normalize the volume of an audio signal. Parameters: audio (torch tensor): Input audio signal array. coeff (float): Target c
soulxpodcast/utils/audio.py:80
↓ 2 callersFunctioncleanup_old_files
清理过期的文件 Args: directory: 要清理的目录 minutes: 文件保留时间(分钟) Returns: int: 清理的文件数量
api/utils.py:133
↓ 2 callersMethodextend_pe
Reset the positional encodings.
soulxpodcast/models/modules/flow_components/upsample_encoder.py:170
↓ 2 callersMethodforward_attention
Compute attention context vector. Args: value (torch.Tensor): Transformed value, size (#batch, n_head, time2, d_k
soulxpodcast/models/modules/flow_components/upsample_encoder.py:424
↓ 2 callersMethodforward_qkv
Transform query, key and value. Args: query (torch.Tensor): Query tensor (#batch, time1, size). key (torch.Tensor): K
soulxpodcast/models/modules/flow_components/upsample_encoder.py:395
↓ 2 callersFunctiongenerate_task_id
生成唯一的任务ID
api/utils.py:19
↓ 2 callersFunctionget_gpu_memory_info
获取GPU内存信息
api/monitor.py:18
↓ 2 callersFunctionget_padding
(kernel_size, dilation=1)
soulxpodcast/models/modules/hifigan_components/layers.py:16
↓ 2 callersMethodinitialize_weights
(self)
soulxpodcast/models/modules/flow_components/estimator.py:662
↓ 2 callersFunctioninitiate_model
(seed, model_path, llm_engine, fp16_flow)
soulxpodcast/utils/infer_utils.py:16
↓ 2 callersFunctionprocess_single_input
(dataset, target_text_list, prompt_wav_list, prompt_text_list, use_dialect_prompt, dialect_prompt_text_list)
soulxpodcast/utils/infer_utils.py:38
↓ 2 callersFunctionsave_upload_file
保存上传的文件到临时目录 Args: upload_file: FastAPI上传文件对象 task_id: 任务ID index: 文件索引 Returns: Path: 保存的文件路径
api/utils.py:24
↓ 2 callersFunctionsubsequent_chunk_mask
Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: size (int): size of mask
soulxpodcast/models/modules/flow_components/upsample_encoder.py:9
↓ 2 callersFunctionvalidate_audio_files
验证音频文件 Args: files: 上传的文件列表 Raises: HTTPException: 如果验证失败
api/utils.py:58
↓ 2 callersFunctionvalidate_dialogue_format
验证对话文本格式 Args: dialogue_text: 对话文本 num_speakers: 说话人数量 Returns: Tuple[bool, str]: (是否有效, 错误信息)
api/utils.py:93
↓ 1 callersMethod__init__
(self, in_channels=320, cfm_params=CfmParams(), n_spks=1, spk_emb_dim=80, estimator: torch.nn.Module = None)
soulxpodcast/models/modules/flow.py:23
↓ 1 callersMethod__init__
( self, in_channels: int = 80, base_channels: int = 512, nb_ha
soulxpodcast/models/modules/hifigan.py:80
↓ 1 callersMethod_f02sine
f0_values: (batchsize, length, dim) where dim indicates fundamental tone and overtones
soulxpodcast/models/modules/hifigan_components/layers.py:305
↓ 1 callersMethod_f02uv
(self, f0)
soulxpodcast/models/modules/hifigan_components/layers.py:179
↓ 1 callersMethod_istft
(self, magnitude, phase)
soulxpodcast/models/modules/hifigan.py:195
↓ 1 callersMethod_load_json_file
Load data from a JSON file
soulxpodcast/config.py:79
↓ 1 callersMethod_load_model
加载模型
api/service.py:49
↓ 1 callersMethod_process_task
处理单个任务
api/tasks.py:102
↓ 1 callersMethod_stft
(self, x)
soulxpodcast/models/modules/hifigan.py:187
↓ 1 callersMethod_worker
后台工作线程
api/tasks.py:72
↓ 1 callersFunctioncheck_dialogue_text
(text_list: List[str])
webui.py:229
↓ 1 callersFunctioncheck_prefix
(text)
soulxpodcast/utils/parser.py:11
↓ 1 callersFunctioncleanup_task
()
api/main.py:65
↓ 1 callersFunctionclear_gpu_cache
清理GPU缓存
api/monitor.py:73
↓ 1 callersFunctiondynamic_range_compression_torch
(x, C=1, clip_val=1e-5)
soulxpodcast/utils/audio.py:22
↓ 1 callersFunctiondynamic_range_decompression_torch
(x, C=1)
soulxpodcast/utils/audio.py:26
↓ 1 callersMethodforward
(self, x, mask)
soulxpodcast/models/modules/flow_components/estimator.py:358
↓ 1 callersMethodforward_layers
(self, xs: torch.Tensor, chunk_masks: torch.Tensor, pos_emb: torch.Tensor,
soulxpodcast/models/modules/flow_components/upsample_encoder.py:985
↓ 1 callersMethodforward_up_layers
(self, xs: torch.Tensor, chunk_masks: torch.Tensor, pos_emb: torch.Tensor,
soulxpodcast/models/modules/flow_components/upsample_encoder.py:992
↓ 1 callersMethodfrom_dataclass
Automatically create configuration from any dataclass
soulxpodcast/config.py:97
↓ 1 callersMethodgenerate
生成语音 Args: prompt_audio_paths: 参考音频路径列表 prompt_texts: 参考文本列表 dialogue_text: 对话文本 see
api/service.py:88
↓ 1 callersFunctiongenerate_time_index
Generate a time-based unique key, e.g. '20251023-001'.
soulxpodcast/utils/parser.py:6
↓ 1 callersMethodget_active_task_count
获取活跃任务数量
api/tasks.py:188
↓ 1 callersFunctionget_args
()
webui.py:579
↓ 1 callersFunctionget_system_memory_info
获取系统内存信息
api/monitor.py:32
↓ 1 callersMethodget_task
获取任务信息
api/tasks.py:184
↓ 1 callersFunctioninitiate_model
(config: Config, enable_tn: bool=False)
webui.py:106
↓ 1 callersFunctionload_dialect_prompt_data
加载方言提示文本文件并格式化为嵌套字典。 返回结构: {dialect_key: {display_name: full_text, ...}, ...}
webui.py:26
↓ 1 callersFunctionmain
()
run_api.py:15
↓ 1 callersFunctionmain
()
api/test_client.py:224
↓ 1 callersFunctionmel_spectrogram
(y, n_fft=1920, num_mels=80, sampling_rate=24000, hop_size=480, win_size=1920, fmin=0, fma
soulxpodcast/utils/audio.py:44
↓ 1 callersFunctionmonitor_memory
持续监控内存使用
api/monitor.py:43
↓ 1 callersMethodoutput_size
(self)
soulxpodcast/models/modules/flow_components/upsample_encoder.py:920
↓ 1 callersFunctionparse_dialogue_text
解析对话文本为列表格式 Args: dialogue_text: 原始对话文本 num_speakers: 说话人数量 Returns: List[str]: 分段的对话列表,每段包含说话人标记
api/utils.py:177
↓ 1 callersMethodposition_encoding
For getting encoding in a streaming fashion Attention!!!!! we apply dropout only once at the whole utterance level in a none
soulxpodcast/models/modules/flow_components/upsample_encoder.py:218
↓ 1 callersFunctionprocess_single
(target_text_list, prompt_wav_list, prompt_text_list, use_dialect_prompt, dialect_prompt_text)
webui.py:242
↓ 1 callersMethodrel_shift
Compute relative positional encoding. Args: x (torch.Tensor): Input tensor (batch, head, time1, 2*time1-1). time1 mea
soulxpodcast/models/modules/flow_components/upsample_encoder.py:566
↓ 1 callersFunctionrender_interface
()
webui.py:364
↓ 1 callersFunctionrun_inference
( inputs: dict, model_path: str, output_path: str, llm_engine: str = "hf", fp16_flow: bool
cli/podcast.py:14
↓ 1 callersFunctionrun_inference
( inputs: dict, model_path: str, output_path: str, llm_engine: str = "hf", fp16_flow: bool
cli/tts.py:14
↓ 1 callersFunctionset_all_random_seed
(seed)
soulxpodcast/utils/commons.py:6
↓ 1 callersMethodshutdown
关闭任务管理器
api/tasks.py:195
↓ 1 callersMethodsolve_euler
Fixed euler solver for ODEs. Args: x (torch.Tensor): random noise t_span (torch.Tensor): n_timesteps interpol
soulxpodcast/models/modules/flow.py:66
↓ 1 callersFunctionspectral_normalize_torch
(magnitudes)
soulxpodcast/utils/audio.py:30
↓ 1 callersMethodstart_workers
启动后台工作线程
api/tasks.py:62
↓ 1 callersFunctiontest_async
测试异步生成
api/test_client.py:112
↓ 1 callersFunctiontest_health
测试健康检查
api/test_client.py:202
↓ 1 callersFunctiontest_sync_multi_speaker
测试同步生成 - 多说话人
api/test_client.py:58
↓ 1 callersFunctiontest_sync_single_speaker
测试同步生成 - 单说话人
api/test_client.py:14
↓ 1 callersMethodvalidate_llm_engine
验证LLM引擎配置
api/config.py:21
Method__getitem__
(self, idx)
soulxpodcast/utils/dataloader.py:82
Method__init__
(self, **kwargs)
soulxpodcast/config.py:90
Method__init__
(self, text_tokenizer, data_list, model_config: Config)
soulxpodcast/utils/dataloader.py:26
Method__init__
(self, text_tokenizer, data_list, model_config: Config)
soulxpodcast/utils/dataloader.py:171
Method__init__
(self, model, **kwargs)
soulxpodcast/engine/llm_engine.py:25
Method__init__
(self, model, **kwargs)
soulxpodcast/engine/llm_engine.py:80
Method__init__
(self, config: Config = None)
soulxpodcast/models/soulxpodcast.py:21
Method__init__
( self, input_size: int = 512, output_size: int = 80, spk_embed_dim: int = 192
soulxpodcast/models/modules/flow.py:129
Method__init__
(self, num_class: int = 1, in_channels: int = 80, cond_chan
soulxpodcast/models/modules/hifigan.py:37
Method__init__
(self, d_model: int, max_len: int = 5000)
soulxpodcast/models/modules/flow_components/upsample_encoder.py:163
Method__init__
(self, idim: int, odim: int, pos_enc_class: torch.nn.Module)
soulxpodcast/models/modules/flow_components/upsample_encoder.py:261
Method__init__
(self, channels: int, pre_lookahead_len: int = 1)
soulxpodcast/models/modules/flow_components/upsample_encoder.py:330
Method__init__
(self, n_head: int, n_feat: int, dropout_rate: float,
soulxpodcast/models/modules/flow_components/upsample_encoder.py:379
Method__init__
(self, n_head: int, n_feat: int, dropout_rate: float,
soulxpodcast/models/modules/flow_components/upsample_encoder.py:551
next →1–100 of 193, ranked by callers