Functions — github.com/ASLP-lab/SenSE

CodeHub MCPcopy Index your code

hub / github.com/ASLP-lab/SenSE / functions

Functions360 in github.com/ASLP-lab/SenSE

⨍ Functions 360◇ Types & classes 78

↓ 1 callersFunctioninitialize_asr_pipeline

(device: str = device, dtype=None)

src/sense/infer/utils_infer.py:156

↓ 1 callersMethodinitialize_weights

src/sense/model/backbones/mmdit.py:130

↓ 1 callersMethodinitialize_weights

src/sense/model/backbones/dit.py:163

↓ 1 callersFunctionis_csv_wavs_format

(input_dataset_dir)

src/sense/train/datasets/prepare_csv_wavs.py:29

↓ 1 callersFunctionload_asr_model

(lang, ckpt_dir="")

src/sense/eval/utils_eval.py:382

↓ 1 callersMethodload_checkpoint

src/sense/model/trainer_llm.py:178

↓ 1 callersMethodload_checkpoint

src/sense/model/trainer_cfm.py:184

↓ 1 callersFunctionload_noise_list

Load noise file list from .scp file.

src/sense/train/noisy_datasets/process_one_file.py:22

↓ 1 callersFunctionload_token_data

加载token文件，返回一个key->code的字典

src/sense/train/datasets/prepare_custom_data.py:8

↓ 1 callersFunctionload_token_data

加载合并的token文件数据 Args: token_file_path (str): token文件路径 Returns: dict: 以key为索引的token字典

src/sense/train/datasets/prepare_libritts_noisy_token.py:17

↓ 1 callersFunctionload_token_data

加载合并的token文件数据 Args: token_file_path (str): token文件路径 Returns: dict: 以key为索引的token字典

src/sense/train/datasets/prepare_emilia_se.py:23

↓ 1 callersFunctionmain

src/sense/eval/eval_dnsmos_oval.py:81

↓ 1 callersFunctionmain

src/sense/eval/eval_infer_batch.py:37

↓ 1 callersFunctionmain

src/sense/eval/eval_dnsmos.py:93

↓ 1 callersFunctionmain

src/sense/eval/eval_utmos.py:10

↓ 1 callersFunctionmain

src/sense/eval/eval_infer_cfm_batch.py:41

↓ 1 callersFunctionmain

src/sense/eval/eval_metrics.py:32

↓ 1 callersFunctionmain

src/sense/infer/infer.py:158

↓ 1 callersFunctionmain

src/sense/infer/infer_llm.py:158

↓ 1 callersFunctionmain

src/sense/infer/infer_cfm.py:212

↓ 1 callersFunctionmain

src/sense/train/train_llm.py:18

↓ 1 callersFunctionmain

src/sense/train/train_cfm.py:18

↓ 1 callersFunctionmain

src/sense/train/finetune.py:19

↓ 1 callersFunctionmain

src/sense/train/noisy_datasets/prepare_libritts_noisy.py:252

↓ 1 callersFunctionmain

src/sense/train/noisy_datasets/run_libritts_noisy.py:9

↓ 1 callersFunctionmain

src/sense/train/noisy_datasets/test_libritts_noisy.py:94

↓ 1 callersFunctionmain

src/sense/train/noisy_datasets/process_one_file.py:156

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_noise.py:141

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_custom_data.py:94

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_emilia_v2.py:44

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_libritts_noisy.py:36

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_libritts.py:34

↓ 1 callersFunctionmain

src/sense/train/datasets/modify_audio_paths.py:79

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_rir.py:175

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_libritts_noisy_token.py:95

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_emilia.py:147

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_wenetspeech4tts.py:49

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_emilia_se.py:184

↓ 1 callersFunctionmain

src/sense/train/datasets/prepare_ljspeech.py:16

↓ 1 callersFunctionmask_from_start_end_indices

(seq_len: int["b"], start: int["b"], end: int["b"])

src/sense/model/utils.py:49

↓ 1 callersFunctionmodify_audio_paths

修改Arrow文件中的音频路径前缀 Args: input_arrow_path: 输入的raw.arrow文件路径 output_arrow_path: 输出的raw.arrow文件路径 old_prefix: 需要替换的

src/sense/train/datasets/modify_audio_paths.py:16

↓ 1 callersFunctionmodify_libritts_noisy_paths

修改LibriTTS_noisy数据集中的音频路径前缀

src/sense/train/datasets/modify_paths_example.py:14

↓ 1 callersMethodpesq_metric

(self, ref, inf, fs=8000)

src/sense/eval/intrusive_se_metrics.py:32

↓ 1 callersFunctionprepare_and_save_set

(inp_dir, out_dir, is_finetune: bool = True, num_workers: int = None)

src/sense/train/datasets/prepare_csv_wavs.py:238

↓ 1 callersFunctionprepare_csv_wavs_dir

(input_dir, num_workers=None)

src/sense/train/datasets/prepare_csv_wavs.py:92

↓ 1 callersFunctionpreprocess_ref_audio_text

(ref_audio_orig, ref_text, tokenizer, show_info=print)

src/sense/infer/utils_infer.py:304

↓ 1 callersFunctionprocess_audio_file

Process a single audio file to create noisy version

src/sense/train/noisy_datasets/process_one_file.py:96

↓ 1 callersFunctionprocess_batch

src/sense/infer/utils_infer.py:502

↓ 1 callersFunctionprocess_custom_jsonl

处理自定义 jsonl 文件，返回处理结果

src/sense/train/datasets/prepare_custom_data.py:34

↓ 1 callersFunctionprocess_rir_scp_files

处理多个RIR SCP文件 Args: scp_files (list): SCP文件路径列表 Returns: list: 合并后的音频信息列表 list: 合并后的时长列表

src/sense/train/datasets/prepare_rir.py:95

↓ 1 callersFunctionprocess_scp_files

处理多个SCP文件 Args: scp_files (list): SCP文件路径列表 Returns: list: 合并后的音频信息列表 list: 合并后的时长列表

src/sense/train/datasets/prepare_noise.py:85

↓ 1 callersFunctionrandom_amplitude_scaling

Normalize and then scale multiple audio tensors by the same random factor. Args: *audios: input torch tensors (1D or 2D) sca

src/sense/model/dataset.py:57

↓ 1 callersFunctionread_audio_text_pairs

(csv_file_path)

src/sense/train/datasets/prepare_csv_wavs.py:189

↓ 1 callersFunctionread_rir_scp_file

读取RIR .scp文件，解析音频路径、时长和采样率信息 Args: scp_file_path (str): .scp文件路径 Returns: list: 包含音频路径和时长的字典列表 list

src/sense/train/datasets/prepare_rir.py:18

↓ 1 callersFunctionread_scp_file

读取.scp文件，解析音频路径和时长信息 Args: scp_file_path (str): .scp文件路径 Returns: list: 包含音频路径和时长的字典列表 list: 时长列表

src/sense/train/datasets/prepare_noise.py:17

↓ 1 callersFunctionremove_silence_edges

(audio, silence_threshold=-42)

src/sense/infer/utils_infer.py:285

↓ 1 callersMethodreset_parameters

src/sense/model/modules2.py:393

↓ 1 callersFunctionsave_dataset

保存为 arrow 和 duration.json 文件

src/sense/train/datasets/prepare_custom_data.py:76

↓ 1 callersFunctionsave_prepped_dataset

(out_dir, result, duration_list, text_vocab_set, is_finetune)

src/sense/train/datasets/prepare_csv_wavs.py:206

↓ 1 callersFunctionsave_rir_to_arrow

保存RIR数据到Arrow格式文件 Args: result (list): 音频信息列表 duration_list (list): 时长列表 save_dir (str): 保存目录 dataset_na

src/sense/train/datasets/prepare_rir.py:119

↓ 1 callersFunctionsave_to_arrow

保存数据到Arrow格式文件 Args: result (list): 音频信息列表 duration_list (list): 时长列表 save_dir (str): 保存目录 dataset_name

src/sense/train/datasets/prepare_noise.py:109

↓ 1 callersMethodscore

Args: gt_wav (np.ndarray): Ground truth waveform (T,). gen_wav (np.ndarray): Generated waveform (T,). Returns

src/sense/eval/SpeechBERTScore/speechbertscore.py:87

( self, attn: Attention, x: float["b n d"], # noised input x # noqa: F722 ma

src/sense/model/modules.py:573

( self, attn: Attention, x: float["b n d"], # noised input x # noqa: F722 c:

src/sense/model/modules.py:645

(self, fpath, sampling_rate)

src/sense/eval/eval_dnsmos_oval.py:40

(self, fpath, sampling_rate)

src/sense/eval/eval_dnsmos.py:40

(self, ref_path, inf_path, metrics=METRICS)

src/sense/eval/intrusive_se_metrics.py:51

Method__getitem__

src/sense/model/dataset.py:136

Method__getitem__

src/sense/model/dataset.py:227

( self, model: LLM_LLaMA, epochs, learning_rate, num_warmup_updates=20

src/sense/model/trainer_llm.py:26

(self, deconstruct_idx=None)

src/sense/model/modules2.py:19

(self, *args, **kwargs)

src/sense/model/modules2.py:30

(self, *args, **kwargs)

src/sense/model/modules2.py:45

(self, kernel_size, causal=False)

src/sense/model/modules2.py:72

(self, input_dim, output_dim, glu_type="sigmoid", bias_in_glu=True)

src/sense/model/modules2.py:99

( self, embed_dim, num_heads, kdim=None, vdim=None

src/sense/model/modules2.py:307

( self, transformer: nn.Module, sigma=0.0, odeint_kwargs: dict = dict(

src/sense/model/cfm.py:33

( self, hf_dataset: Dataset, target_sample_rate=24_000, n_mel_channels=100,

src/sense/model/dataset.py:104

( self, custom_dataset: Dataset, durations=None, target_sample_rate=24_000,

src/sense/model/dataset.py:167

( self, sampler: Sampler[int], frames_threshold: int, max_samples=0, random_seed=None, drop_residual:

src/sense/model/dataset.py:298

( self, n_fft=1024, hop_length=256, win_length=1024, n_mel_channels=10

src/sense/model/modules.py:230

src/sense/model/modules.py:281

(self, dim, kernel_size=31, groups=16)

src/sense/model/modules.py:299

( self, dim: int, intermediate_dim: int, dilation: int = 1, )

src/sense/model/modules.py:373

(self, dim: int, eps: float)

src/sense/model/modules.py:407

src/sense/model/modules.py:433

src/sense/model/modules.py:454

(self, dim, dim_out=None, mult=4, dropout=0.0, approximate: str = "none")

src/sense/model/modules.py:474

( self, processor: JointAttnProcessor | AttnProcessor, dim: int, heads: int =

src/sense/model/modules.py:492

( self, pe_attn_head: int | None = None, # number of attention head to apply rope, None for a

src/sense/model/modules.py:567

src/sense/model/modules.py:642

(self, dim, heads, dim_head, ff_mult=4, dropout=0.1, qk_norm=None, pe_attn_head=None)

src/sense/model/modules.py:742

( self, dim, heads, dim_head, ff_mult=4, dropout=0.1, context_dim=None, context_pre_only=False, qk_nor

src/sense/model/modules.py:788

(self, dim, freq_embed_dim=256)

src/sense/model/modules.py:853

( self, model: CFM, epochs, learning_rate, num_warmup_updates=20000,

src/sense/model/trainer_cfm.py:27

(self, ds_rate, encoder_dim, llm_dim)

src/sense/model/projector.py:5

(self, encoder_dim, llm_dim, qformer_layers, query_len)

src/sense/model/projector.py:51

( self, encoder_name: str = "conformer", freeze_encoder: bool = False, encoder

src/sense/model/llama_llm.py:19

(self, out_dim, text_num_embeds, mask_padding=True)

src/sense/model/backbones/mmdit.py:30

(self, in_dim, out_dim)

src/sense/model/backbones/mmdit.py:67

← previous next →101–200 of 360, ranked by callers