Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` Args: input_shape (`tuple(int)` or `list(int)` o

Initialize the weights.

Initialize the weights

Initialize the weights

Creates a non-causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape `(batch_size, key_value_lengt

Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape `(batch_size, key_value_length)`

Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel

Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel

Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel

same output as PIL.ImageOps.autocontrast

This method forwards all its arguments to CLIPTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please refer to the docstring of

same output as PIL.ImageEnhance.Contrast

r""" Args: labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked l

same output as PIL.ImageEnhance.Color

same output as PIL.ImageEnhance.Contrast

Setup the optimizer. We provide a reasonable default that works well. If you want to use something else, you can pass a tuple in the

A decorator to check if some arguments are deprecate and try to replace deprecate src_arg_name to dst_arg_name. Args: name_dict(dict):

A decorator to check if some arguments are deprecate and try to replace deprecate src_arg_name to dst_arg_name. Args: name_dict(dict):

Detokenize the generated tokens.

Detokenize the generated tokens.

Disable the redundant torch default initialization to accelerate model creation.