args: hyp: list of hypothesis output: list of hypothesis (string)>

Evaluation args: model: Model object test_loader: DataLoader object

For masking out the padding part of key sequence.

Get word segments and tokenize the sequence for selected language We cannot run two different languages on stanford core nlp, will be very sl

Check if the character is Chinese args: cc: char output: boolean

Loading discriminator args: load_path: string

Loading model args: load_path: string

Picks tempo and gain uniformly, applies it to the utterance by using sox utility. Returns the augmented utterance.

args: hyp: list of hypothesis output: list of hypothesis (string)>

Takes samples (a batch) and returns Inputs: samples, start_letter, cuda - samples: batch_size x seq_len (Tensor with a sample in

Implement `lrate` above

Remove english and chinese punctuation except hypen/dash, and full stop. Also fix some typos and encoding issues args: seq: Strin

Remove special characters from the corpus args: seq: String output: seq: String

Wraps hidden states in new Tensors, to detach them from their history.

Parses audio file into spectrogram with optional normalization and various augmentations :param audio_conf: Dictionary containing t

Dataset that loads tensors via a csv containing file paths to audio files and transcripts separated by a comma. Each new line is a

Dataset that loads tensors via a csv containing file paths to audio files and transcripts separated by a comma. Each new line is a

Samples batches assuming they are in order of size to batch similarly sized samples together.

Converts an index (integer) in a token (string/unicode) using the vocab.

Converts a token (str/unicode) in an id using the vocab.

Training args: model: Model object train_loader: DataLoader object of the training set valid

Converts a sequence of tokens (string) in a single string.

Beam search, decode nbest utterances args: encoder_padded_outputs: B x T x H beam_size: int nbest: int outp

Greedy search, decode 1-best utterance args: encoder_padded_outputs: B x T x H output: batch_ids_nbest_hyps: list of

args: hyp: list of hypothesis output: list of hypothesis (string)>

Returns a dictionary containing the encoded sequence or sequence pair and additional informations: the mask for sequence classifica