See `pad_for_conv1d`.

Return n_q based on specified target bandwidth.

Tiny wrapper around F.pad, just to allow for reflect padding on small input. If this is the case, we insert extra 0 padding to the right before th

For getting encoding in a streaming fashion Attention!!!!! we apply dropout only once at the whole utterance level in a none

Pull a symbol, reading as many bits from the stream as required. This returns `None` when the stream has been exhausted. Args:

Given an audio and the PaSST model, return the probabilities of each AudioSet class. Audio is converted to mono at 32kHz. PaSST model i

Sample data according to rank/world_size/num_workers Args: data(List): input data list Returns:

Create mask for subsequent steps (size, size) with chunk size, this is for streaming encoder Args: size (int): size of mask

Tie or clone module weights (between word_emb and output_layer) depending of whether we are using TorchScript or not

Decode text to chars or BPE Inplace operation Args: data: Iterable[{key, wav, txt, sample_rate}] Returns:

Trims the audio from the beginning to the first occurrence of silence. Parameters: waveform (Tensor): The waveform data to the input

Remove padding from x, handling properly zero padding. Only for 1d!

Args: feature_extractor (FeatureExtractor): An instance of FeatureExtractor to extract features from audio signals. b

Construct an MultiHeadedAttention object.

Construct an EncoderLayer object.

Simple const lr policy

Simple const lr policy

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

Adapted from: https://github.com/mseitzer/pytorch-fid/blob/master/src/pytorch_fid/fid_score.py Adapted from: https://github.com/gudgud96/frec

Convert audio wave form to a specified lossy format: mp3, ogg, flac Args: wav (torch.Tensor): Input wav tensor. sr (int): Samplin

Computes diffusion loss Args: x1 (torch.Tensor): Target shape: (batch_size, n_feats, mo) mask (torch.

Create time embedding for the given positions, target dimension `dim`.

Method to decode audio waveform from already calculated features. The features input is passed through the backbone and the head to r

Method to decode audio waveform from already calculated features. The features input is passed through the backbone and the head to r

Detects the first occurrence of silence in the audio. Parameters: audio (Tensor): The audio waveform. sample_rate (int): The

Dynamic batch the data until the total frames in batch reach `max_frames_in_batch` Args: data: Iterable[{key, feat, labe

Class method to create a new Vocos model instance from a pre-trained model stored in the Hugging Face model hub.

Flush the remaining information to the stream.

Forward diffusion Args: mu (torch.Tensor): output of encoder shape: (batch_size, n_feats, mel_timesteps)

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