config.json

{
    "tokenizer_name": "word_level",
    "apply_pca": 256,
    "apply_zipf": true,
    "normalize": false,
    "hidden_dim": 256,
    "seq_length": 1000000
}