Quantize the weights of the model from float32 to in8 to allow very efficient inference on modern CPU Args: onnx_model_path: Path to

Remove a callback from the current list of :class:`~transformer.TrainerCallback`. Args: callback (:obj:`type` or :class:`

Retrieves documents for specified ``question_hidden_states``. The main process, which has the access to the index stored in memory, gathers q

Takes input text, generates output, and then using reference calculates the BLEU scores. The results are saved to a file and returned to th

Save the current state. Only by the master process.

Save the content of this instance in JSON format inside :obj:`json_path`.

Rescores a prediction based on penalty functions

Reset the special tokens to the source lang setting. No prefix and suffix=[eos, cur_lang_code].

Reset the special tokens to the source lang setting. No prefix and suffix=[eos, cur_lang_code].

Filter a distribution of logits using top-k and/or nucleus (top-p) filtering Args: logits: logits distribution shape (batch size, vo

Tie the weights between the input embeddings and the output embeddings. If the :obj:`torchscript` flag is set in the configuration,

Serializes this instance while replace `Enum` by their values (for JSON serialization support).

Sanitized serialization to use with TensorBoard’s hparams

Tokenizes a piece of text into characters. For example, :obj:`input = "apple""` wil return as output :obj:`["a", "p", "p", "l", "e"]

Call HF API to know "whoami"

Do not add eos again if user already added it.

If tensors not close, or a and b arent both tensors, raise a nice Assertion error.

convert into overlapping chunks. Chunk size = 2w, overlap size = w

Computes global attention mask by putting attention on all tokens before `sep_token_id` if `before_sep_token is True` else after `sep_token_i

Compute softmax probability over raw logits.

Creates examples for the training and dev sets.

Evaluation called for both Val and Test

Internal framework specific forward dispatching Args: inputs: dict holding all the keyword arguments for required by the

expand dims of idxs and vectors for all hashes and gather

Original Implementation of the GELU activation function in Google BERT repo when initially created. For information: OpenAI GPT's GELU is sli

Get the n-best logits from a list.

Return the shape of the first array found in the nested struct `arrays`.

Returns the indentation in the first line of t