一、手工标注
1、创建数据集
2、打分
3、指定数据集
并且可以在Expected Output中输入期望的输出
二、上传数据集
1、编写代码
from langfuse import Langfuse
import os
# 加载 .env 到环境变量
from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv())
import json
from langfuse.model import CreateDatasetRequest, CreateDatasetItemRequest
from tqdm import tqdm
data = []
with open('my_langfusedataset.jsonl', 'r', encoding='utf-8') as fp:
for line in fp:
example = json.loads(line.strip())
item = {
"input": {
"outlines": example["outlines"],
"user_input": example["user_input"]
},
"expected_output": example["label"]
}
data.append(item)
# init
langfuse = Langfuse()
for item in tqdm(data[:50]):
langfuse.create_dataset_item(
dataset_name="myDataset1",
input=item["input"],
expected_output=item["expected_output"]
)