Assume the first argument is the name of the argument.

r"""Get an item from registry with key 'name' Args: name (string): Key whose value needs to be retrieved. default: If

Build a pretrained model from default configuration file, specified by model_type. Args: - model_type (str): model type,

Get the attributions for (part of) the response. Arguments: start_idx (int, optional): Start index of th

r"""Register a path to registry with key 'name' Args: name: Key with which the path will be registered. Usage:

Download a file from a URL and cache it locally. If the file already exists, it is not downloaded again. If distributed, only the main proces

Compute sequence-level outputs from token-level logit-probabilities.

Performs all_gather operation on the provided tensors. *** Warning ***: torch.distributed.all_gather has no gradient.

Get a represention of the source corresponding to a given index.

Create the directory if it does not exist.

r"""Register an item to registry with key 'name' Args: name: Key with which the item will be registered. Usage::

Save the checkpoint at the current epoch.

An inner training loop compatible with both epoch-based and iter-based training. When using epoch-based, training stops after one ep

Download a file from a url and place it in root. Args: url (str): URL to download file from root (str): Directory to place downloa

Evaluate the model on a given split. Args: split_name (str): name of the split to evaluate on. cur_epoch (in

Makes broadcastable attention and causal masks so that future and masked tokens are ignored. Arguments: attention_mask (

Load from a finetuned checkpoint. This should expect no mismatch in the model keys and the checkpoint keys.

Rotates half the hidden dims of the input.

Split text into parts and cache the parts and separators.