---
dataset_info:
  features:
  - name: index
    dtype: int32
  - name: token_count
    dtype: int32
  - name: top_counts
    list: int8
  - name: top_indices
    list: int32
  - name: top_logits
    list: float16
  - name: lses
    list: float32
  - name: sampled_indices
    list:
      list: int32
      length: 12
  - name: sampled_logits
    list:
      list: float16
      length: 12
  splits:
  - name: train
    num_bytes: 9765066162
    num_examples: 103357
  download_size: 5683181466
  dataset_size: 9765066162
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---