tokenized/metadata.json

{
  "vocab_size": 50277,
  "sequence_length": 2049,
  "num_train": 819200,
  "num_eval": 102400,
  "num_test": 102400,
  "total_samples": 1024000,
  "dtype": "uint16",
  "seed": 42
}