metadata
license: cc-by-sa-3.0
source_datasets:
- databricks/databricks-dolly-15k
Conversion of databricks/databricks-dolly-15k dataset to be used in pretraining.
Python code used for conversion:
from datasets import load_dataset
import pandas
dataset = load_dataset("databricks/databricks-dolly-15k", split="train")
def format(columns):
instruction = columns["instruction"].strip()
answer = columns["response"].strip()
return f"{instruction}\n\n{answer}"
pandas.DataFrame({"text": [format(columns) for columns in dataset]}).to_csv("train.csv", index=False)