added name to training sample conversion for helpsteer and oasst

This commit is contained in:
wea_ondara
2024-04-24 19:26:22 +02:00
parent f4aefb3278
commit 9cafd8b1e9
2 changed files with 19 additions and 7 deletions

View File

@@ -1,7 +1,9 @@
import json
import os
import random
this_dir = os.path.dirname(os.path.abspath(__file__))
user_names = ['Adam', 'Alice', 'Anne', 'Bob', 'Charlie', 'Cody', 'Corinna', 'Cynthia', 'Fred', 'Grace', 'Jane', 'Paul',
'Rachel', 'Ramesh']
def mkdir(path):
@@ -9,6 +11,7 @@ def mkdir(path):
os.mkdir(path)
this_dir = os.path.dirname(os.path.abspath(__file__))
mkdir(this_dir + '/../../data')
mkdir(this_dir + '/../../data/helpsteer')
@@ -18,8 +21,10 @@ for filename in ['train.jsonl', 'validation.jsonl']:
role_dict = {'prompt': 'user', 'response': 'assistant'}
lines = [json.loads(line) for line in lines]
conversations = [{'messages': [{'role': 'user', 'content': line['prompt']},
{'role': 'assistant', 'content': line['response']}]} for line in lines]
conversations = [{'messages': [
{'role': 'user', 'name': user_names[random.randint(0, len(user_names) - 1)], 'content': line['prompt']},
{'role': 'assistant', 'name': 'assistant', 'content': line['response']}]
} for line in lines]
print(conversations[0])