Args: input_size (int): input dim output_size (int): dimension of attention attention_heads (int): the nu

prompt_patern1,prompt,history, wav, prompt_patern2,txt,answer_wav, historcy_batch的内容是： [ big_embed, [

Simple const lr policy

Simple const lr policy

Simple const lr policy

Sample `num_negatives` vectors from feature vectors.

Prepad blank for transducer predictor Args: ys_pad (torch.Tensor): batch of padded target sequences (B, Lmax) blank (int): index

Apply optional mask for encoder. Args: xs (torch.Tensor): padded input, (B, L, D), L for max length mask (torch.Tensor): mask fo

Args: ctc_prefix_results(List[DecodeResult]): ctc prefix beam search results

Returns: List[List[List[int]]]: nbest result for each utterance

Cross validation on

得到第一个<>标签的内容，并将其转为大写 >>> input : "<tag1>content1<tag2>content2<tag3>content3" >>> output: "<TAG1>" :param upper: if false, return the

r""" Estimate the gradient of the model on the given dataset

Reset the positional encodings.

ctc forced alignment. Args: torch.Tensor ctc_probs: hidden state sequence, 2d tensor (T, D) torch.Tensor y: id sequence tensor 1d

Forward just one chunk Args: xs (torch.Tensor): chunk input, with shape (b=1, time, mel-dim), where `time == (ch

Forward just one chunk Args: xs (torch.Tensor): chunk input, with shape (b=1, time, mel-dim), where `time == (ch

Transform query, key and value. Args: query (torch.Tensor): Query tensor (#batch, time1, size). key (torch.Tensor): K

将 item_str 插入到 lst 的第 position 个位置（1-based）， 若 lst 长度不足则以 "-1" 填充至目标长度后再插入。

Convert a lexicon to an FST (in k2 format). Args: lexicon: The input lexicon. See also :func:`read_lexicon` token2id:

Generate a db-domain ploynomial distortion function f(x) = a * x^m * (1-x)^n + x Args: conf: a dict {'a': #int, 'm': #int, 'n':

sentence, ignore_words are both in unicode

sentence, ignore_words are both in unicode