Runs basic whitespace cleaning and splitting on a piece of text.

force one of token_ids to be generated by setting prob of all other tokens to 0 (logprob=-float("inf"))

Helper that gets a list of the model's attention modules. Each module has a `bias` buffer used for causal masking. The Prefix LM conv

Return a process group based on gloo backend, containing all the ranks The result is cached.

Returns: list[int]: size of the tensor, on each rank Tensor: padded tensor that has the max size

Replace LlamaAttention modules with cached AdaptedAttention modules.

Tokenize a list of strings.

Tokenize a list of strings.

Run all_gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object group: a torch process

Apply transformation `deltas` (dx, dy, dw, dh) to `boxes`. Args: deltas (Tensor): transformation deltas of shape (N, k*4)

Applies layer norm to the inputs.

Applies layer norm to the inputs.

Returns list of utf-8 byte and a corresponding list of unicode strings. The reversible bpe codes work on unicode strings. This means you

Returns list of utf-8 byte and a corresponding list of unicode strings. The reversible bpe codes work on unicode strings. This means you

Compute binned relative position bias

Deletes an existing adapter. Args: adapter_name (str): Name of the adapter to be deleted.

Dump the result to a GQA-challenge submittable json file. GQA json file submission requirement: results = [result]

train_features = [convert_example_to_features(example, self.max_seq_length, self.tokenizer) for example in examples]

r""" Loads a configuration file from a json file. Args: path_json_file (`str`): The path to the json file