Combine all balanced and unbalanced indexes hdf5s to a single hdf5. This combined indexes hdf5 is used for training with full data (~20k balanced

Computes where to split an utterance waveform and its corresponding mel spectrogram to obtain partial utterances of <partial_utterance_n_fra

r"""Compute similarity between text and audio embeddings

Compute the value. When current unit is larger than previous unit, current unit * all previous units will be used as all previous uni

CONVERT F0 TO CONTINUOUS F0 Args: f0 (ndarray): original f0 sequence with the shape (T) Return: (ndarray): continuous f0 with

Convert applicable model parameters to fp16

一百八 to 一百八十 一亿一千三百万 to 一亿 一千万 三百万

cosine schedule as proposed in https://openreview.net/forum?id=-NEXDKk8gZ

cosine schedule as proposed in https://openreview.net/forum?id=-NEXDKk8gZ

Create indexes a for dataloader to read for training. When users have a new task and their own data, they need to create similar indexes. The

Create black list. Black list is a list of audio ids that will be skipped in training.

Transform a sequence of int ids into a their string versions. This method supports transforming individual input/output ids to their

Decoding operation of timestep t

decode_with_timestamps Decodes the predicted label array (2d) into a list of [(Labelname, onset, offset), ...] :param encoder: Encoder du

Design prototype filter for PQMF. This method is based on `A Kaiser window approach for the design of prototype filters of cosine modulated f

Download videos and extract audio in wav format.

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). This is the same as the DropConnect impl I created for Ef

Computes Dynamic Time Warping (DTW) of two sequences. :param array x: N1*M array :param array y: N2*M array :param func dist: distan

Forward data to a model. Args: model: object generator: object return_input: bool return_target: bool Returns:

:param input: audio signal to be warped (B x 2 x T) :param warpfield: the corresponding warpfield (B x 2 x T) :return: the wa

Apply layer normalization. :param torch.Tensor x: input tensor :return: layer normalized tensor :rtype torch.Tensor

:param x: [B, T, C] :param padding_mask: [B, T] :return: [B, T, C] or [L, B, T, C]

r"""Load list of audio files and return a audio embeddings