def tokenize_dataset()

in step5_data_parallel_naive/dataloader.py [0:0]

• 19 lines of code
• 2 McCabe index (conditional complexity)

    def tokenize_dataset(self, dataset, text_column_name, sequence_length, num_proc):
        """Tokenize the dataset and group texts in chunks of sequence_length + 1"""
        tokenizer_func = partial(
            self.tokenizer_group_text,
            tokenizer=self.tokenizer,
            sequence_length=sequence_length
        )

        tokenized_dataset = dataset.map(
            tokenizer_func,
            input_columns=text_column_name,
            remove_columns=dataset.column_names,
            features=Features({
                "input_ids": Sequence(feature=Value(dtype="int64"), length=sequence_length + 1)
            }),
            batched=True,
            num_proc=num_proc,
            load_from_cache_file=True, # Preprocess dataset only once and cache it
            desc=f"Grouping texts in chunks of {sequence_length+1}",
        )

        return tokenized_dataset