metadata
dataset_info:
features:
- name: problem
dtype: string
- name: answer
dtype: string
- name: tokenized_prompt
dtype: string
- name: responses
list: string
- name: rewards
list: float64
- name: mean_reward
dtype: float64
splits:
- name: train
num_bytes: 22329075280
num_examples: 33600
download_size: 7630963646
dataset_size: 22329075280
configs:
- config_name: default
data_files:
- split: train
path: data/train-*