metadata
dataset_info:
features:
- name: video
dtype:
video:
decode: false
- name: model_output
dtype: string
- name: input_ids
sequence:
sequence: int32
- name: attention_mask
sequence:
sequence: int8
- name: pixel_values_videos
sequence:
sequence: float32
- name: video_grid_thw
sequence:
sequence: int64
splits:
- name: train
num_bytes: 3313520906.6296296
num_examples: 64
- name: test
num_bytes: 880243338.3703704
num_examples: 17
download_size: 569005137
dataset_size: 4193764245
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*