Load an audio file and return PCM along with the sample rate

Reorder incremental state. This should be called when the order of the input has changed from the previous time step. A typical use c

Check ETag on S3 object.

Pull a file directly from S3.

Sets the beam size in the decoder and all children.

Helper to set default arguments based on *add_args*.

Set the number of parameters updates.

Setup the task (e.g., load dictionaries).

Return an example's size as a float or tuple.

Return an example's size as a float or tuple. This value is used when filtering a dataset with ``--max-positions``.

embedding is a (T, B, D) tensor padding_mask is a (B, T) tensor or None attn_mask is a (T, T) tensor or None

Return the optimizer's state dict.

Return the optimizer's state dict.

Performs a single optimization step.

Performs a single optimization step.

Suppress printing on the current device. Force printing with `force=True`.

Train the model for one epoch.

Train the model for one epoch.

Task level update when number of update increases. This is called after optimization step and learning rate update of each step

Convert `url` into a hashed filename in a repeatable way. If `etag` is specified, append its hash to the URL's, delimited by a period.

Clears the gradients of all optimized parameters.

Weight-normalized Conv1d layer

Weight-normalized Conv1d layer

Wrap a *model* to support distributed data parallel training. This is similar to the built-in DistributedDataParallel, but allows additi