lm1b

ds = tfds.load('huggingface:lm1b/plain_text')

A benchmark corpus to be used for measuring progress in statistical language modeling. This has almost one billion words in the training data.

{
    "text": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}