Log something with a given level.

Decode the given codes to a reconstructed representation. Args: codes (torch.Tensor): Int tensor of shape [B, K, T] Retu

Encode the given input tensor to quantized representation. Args: x (torch.Tensor): Float tensor of shape [B, C, T] Retur

Context manager to enter streaming mode. Reset streaming state on exit.

Returns a 0. value on the given device without introducing a synchronization point.

Create normalization module for transformer encoder layer. Args: norm_type (str): Normalization method. dim (int): Dimension of t

Given a tensor of codes of shape `[*]`, returns a tensor of shape `[*, D]`, corresponding to the centroids associated to each code index.

torch.multinomial with arbitrary number of dimensions, and number of candidates on the last dimension. Args: input (torch.Tensor): The in

Decode from the discrete codes to continuous latent space.

Downloads the checkpoints from the given repo, along with its config. Extra overrides are possible for each of Moshi, Mimi, or the text token

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

Reset the streaming state.

Projects a batch of waveforms to unquantized latent space. Args: x (torch.Tensor): Float tensor of shape [B, C, T]. Retu

Renormalizes values from `rvq_first` and `rvq_rest` and adds them. This allows correcting statistics that are normalized by the number of qua