Initialize torch.distributed.

List all available args of the current model.

Defines heuristics to log different conditionings. These can be lists of strings (text-to-image), tensors, ints, ...

Fetch args from only --base

Return node and worker info for PyTorch and some distributed environments.

Compute a distinct, deterministic RNG seed for each worker and node.

Quantizes z, returns quantized zhat, same shape as z.

Restore the parameters stored with the `store` method. Useful to validate the model with EMA parameters without affecting the

Round with straight through gradients.

Save the current parameters for restoring later. Args: parameters: Iterable of `torch.nn.Parameter`; the parameters to be

Expand a stream of open tar files into a stream of tar file contents. This returns an iterator over (filename, file_contents).

Iterate over tar file, yielding filename, content pairs for the given tar stream. :param fileobj: byte stream suitable for tarfile :param met

x: (N, T/2 * S, patch_size**3 * C) imgs: (N, T, H, W, C)

Open URLs and yield a stream of url+stream pairs. Args: data: iterator over dict(url=...) handler: exception handler. kw: