metadata
language: en
license: cc-by-3.0
dataset_info:
features:
- name: instruction
dtype: string
- name: context
dtype: string
- name: response
dtype: string
- name: category
dtype: string
splits:
- name: train
num_bytes: 12208856
num_examples: 14863
- name: validation
num_bytes: 117314
num_examples: 151
download_size: 7866269
dataset_size: 12326170
Dataset Card for "dolly-15k"
Summary
This is the dataset supplied by Databricks for training Dolly V2. This set is split 99% training / 1% validation, should you want to set aside some records for evaluation purposes.