metadata
dataset_info:
features:
- name: question
dtype: string
- name: formatted_prompt
dtype: string
- name: answer
dtype: string
- name: answer_clean
dtype: string
- name: prompt_token_ids
list: int64
- name: answer_token_ids
list: int64
- name: teacher_distributions
list:
- name: token_texts
list: string
- name: probs
list: float64
- name: token_ids
list: int64
- name: metadata
struct:
- name: prompt_length
dtype: int64
- name: answer_length
dtype: int64
- name: model
dtype: string
- name: top_k
dtype: int64
- name: finish_reason
dtype: string
splits:
- name: train
num_bytes: 6512113759
num_examples: 20000
download_size: 2879172888
dataset_size: 6512113759
configs:
- config_name: default
data_files:
- split: train
path: data/train-*