metadata
dataset_info:
features:
- name: seq
dtype: string
- name: full_logits
list:
list:
list: float32
- name: encoder_hidden
list:
list:
list: float32
- name: encoder_boundary_probs
list:
list: float32
- name: k_proj
list:
list: float32
- name: q_proj
list:
list: float32
- name: compression_ratio
dtype: float64
- name: seq_label
dtype: string
splits:
- name: train
num_bytes: 4457024376
num_examples: 109
download_size: 1777502205
dataset_size: 4457024376
configs:
- config_name: default
data_files:
- split: train
path: data/train-*