in evolve-instruct/evolve.py [0:0]
def create_seed_prompts(self):
"""
Turn self.seed_data into a list of strings of text self.source_text_list
Each text string can represent as little as a word, or as much as document.
Just has to be representative of some concept or body of text.
:return: None
"""
import os
if isinstance(self.seed_data, str) and os.path.exists(self.seed_data):
data = load_dataset("json", data_files=self.seed_data)
self.seed_text_list = []
for d in data['train']:
s = ""
if isinstance(self.column_names, str):
s = d[self.column_names]
else:
for col in self.column_names:
s += d[col] + "\n"
self.seed_text_list.append(s.strip())
assert self.seed_text_list, "data import failed, got empty list"