Returns the value of `key` if it exists, else `default`.

Adds {name, value} pair to hyperparameters. Args: name: Name of the hyperparameter. value: Value of the hyperparamete

register default config and parse

Override existing hyperparameter values, parsing new values from a string. See parse_values for more detail on the allowed format for values.

Return the hyperparameter values as a Python dictionary. Returns: A dictionary with hyperparameter names as keys. The values are

Generate a square mask for the sequence for mult-head attention. The masked positions are filled with float(1.0). Unmasked positions are

Set params. Args: config: contains the following fifteen optional parameters: 'window_length': Window length in seco

returns word-error-rate calculated as num_err/num_total

argmax for the Conformer CTC model Args: samples: the data source to be decoded hparams: decoding configs ar

convert a sentence to a list of ids, with special tokens added.

Set the value of an existing hyperparameter. This function verifies that the type of the value matches the type of the existing hyper

Accumulate errors and counts

return tf.data.Dataset object

decode interface

load model

for special data prepare carefully: do not change the shape of samples

config property

Stable log add

Read wav from file. Can be called directly without ReadWav class. Examples:: >>> audio_data, sample_rate = read_wav('test.wav', audio_ch

reset the metrics

creates model using configurations in json, load from checkpoint if previous models exist in checkpoint dir

return subvector */

convert a list of ids to a list of symbols

Args: outputs: true labels, shape: [batch, y_steps, feat_dim] ori_lens: scalar Returns:: reshape

CTC prefix beam search inner implementation reference: https://github.com/wenet-e2e/wenet/blob/main/wenet/transformer/asr_model.py. A

do NN LM forward computation, for both train and decode.

Generate a square mask for the sequence. The masked positions are filled with float(1.0). Unmasked positions are filled with float(0.0).

Override existing hyperparameter values, parsing new values from a dictionary. Args: values_dict: Dictionary of name:value pairs.

used for get logit length

for the finished hyps, mask all the selected word to eos Args: preds: shape:[batch_size*beam_size, beam_size] flag:

for the score of finished hyps, mask the first to 0.0 and the others -inf Args: scores: candidate scores at current step s

used for get logit length

get loss

initialize hvd devices, should be called firstly

parse config

Serializes the hyperparameters into JSON. Args: indent: If a non-negative integer, JSON array elements and object members

used for get logit length

Restore n-best avg checkpoint, if 'sort_by_time' is False, the n-best order is sorted by 'sort_by'; If 'sort_by_time' is True, select

Generate a square mask for the sequence for mult-head attention. The masked positions are filled with bool(True). Unmasked positions are

get the wave file length(duration) in ms Args: wave_file: the path of wave file Returns: wav_length: the length(ms) of the w

save cmvn variables determined by variable_list to file Args: variable_list (list): e.g. ["speaker", "mean", "var"]