Spaces:
Running
Running
File size: 8,345 Bytes
a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 7ecbf9d a984ba9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 |
#!/usr/bin/python3
# -*- coding: utf-8 -*-
"""
https://help.aliyun.com/zh/model-studio/qwen-api-reference
https://help.aliyun.com/zh/model-studio/models
https://help.aliyun.com/zh/model-studio/models?spm=a2c4g.11186623.0.i4#d4ccf72f23jh9
https://help.aliyun.com/zh/model-studio/text-generation?spm=a2c4g.11186623.0.0.6b772e068nnT1J#24e54b27d4agt
Deep-Thinking
https://help.aliyun.com/zh/model-studio/deep-thinking?spm=a2c4g.11186623.0.0.56076f58IJd4mP
"""
import argparse
from datetime import datetime
import json
import os
from pathlib import Path
import sys
import time
from zoneinfo import ZoneInfo # Python 3.9+ 自带，无需安装
pwd = os.path.abspath(os.path.dirname(__file__))
sys.path.append(os.path.join(pwd, "../"))
from openai import OpenAI
from project_settings import environment, project_path
def get_args():
parser = argparse.ArgumentParser()
parser.add_argument(
"--model_name",
# default="qwen3-max-2025-09-23",
# default="qwen3-max-preview",
default="qwen-plus-2025-12-01",
# default="qwen-turbo-2025-07-15",
# default="qwen-flash-2025-07-28",
type=str
)
parser.add_argument(
"--eval_dataset_name",
default="agent-nxcloud-zh-375-choice.jsonl",
type=str
)
parser.add_argument(
"--eval_dataset_dir",
default=(project_path / "data/dataset").as_posix(),
type=str
)
parser.add_argument(
"--eval_data_dir",
default=(project_path / "data/eval_data").as_posix(),
type=str
)
parser.add_argument(
"--client",
default="shenzhen_sase",
type=str
)
parser.add_argument(
"--service",
default="aliyun_api_key_bj",
# default="aliyun_api_key_sgp",
type=str
)
parser.add_argument(
"--create_time_str",
default="null",
# default="20251209_104855",
type=str
)
parser.add_argument(
"--interval",
default=1,
type=int
)
args = parser.parse_args()
return args
def conversation_to_str(conversation: list):
conversation_str = ""
for turn in conversation:
role = turn["role"]
content = turn["content"]
row_ = f"{role}: {content}\n"
conversation_str += row_
return conversation_str
def main():
args = get_args()
eval_dataset_dir = Path(args.eval_dataset_dir)
eval_dataset_dir.mkdir(parents=True, exist_ok=True)
eval_data_dir = Path(args.eval_data_dir)
eval_data_dir.mkdir(parents=True, exist_ok=True)
if args.create_time_str == "null":
tz = ZoneInfo("Asia/Shanghai")
now = datetime.now(tz)
create_time_str = now.strftime("%Y%m%d_%H%M%S")
# create_time_str = "20250724_090615"
else:
create_time_str = args.create_time_str
eval_dataset = eval_dataset_dir / args.eval_dataset_name
model_name_ = args.model_name.replace("/", "#")
output_file = eval_data_dir / f"aliyun_nxcloud_choice2/aliyun/{model_name_}/{args.client}/{args.service}/{create_time_str}/{args.eval_dataset_name}"
output_file.parent.mkdir(parents=True, exist_ok=True)
api_key = environment.get(args.service, dtype=str)
if args.service == "aliyun_api_key_bj":
base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
elif args.service == "aliyun_api_key_sgp":
base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
else:
raise AssertionError(f"invalid service: {args.service}")
client = OpenAI(
base_url=base_url,
# Read your Ark API Key from the environment variable.
api_key=api_key
)
total = 0
total_correct = 0
# finished
finished_idx_set = set()
if os.path.exists(output_file.as_posix()):
with open(output_file.as_posix(), "r", encoding="utf-8") as f:
for row in f:
row = json.loads(row)
idx = row["idx"]
total = row["total"]
total_correct = row["total_correct"]
finished_idx_set.add(idx)
print(f"finished count: {len(finished_idx_set)}")
with open(eval_dataset.as_posix(), "r", encoding="utf-8") as fin, open(output_file.as_posix(), "a+", encoding="utf-8") as fout:
for row in fin:
row = json.loads(row)
idx = row["idx"]
system_prompt = row["system_prompt"]
conversation = row["conversation"]
examples = row["examples"]
choices = row["choices"]
response = row["response"]
if idx in finished_idx_set:
continue
# conversation
conversation_str = conversation_to_str(conversation)
examples_str = ""
for example in examples:
conversation_ = example["conversation"]
outputs = example["outputs"]
output = outputs["output"]
explanation = outputs["explanation"]
examples_str += conversation_to_str(conversation_)
output_json = {"Explanation": explanation, "output": output}
output_json_str = json.dumps(output_json, ensure_ascii=False)
examples_str += f"\nOutput: {output_json_str}\n"
# print(examples_str)
choices_str = ""
for choice in choices:
condition = choice["condition"]
choice_letter = choice["choice_letter"]
row_ = f"{condition}, output: {choice_letter}\n"
choices_str += row_
# choices_str += "\nRemember to output ONLY the corresponding letter.\nYour output is:"
choices_str += "\nPlease use only 10-15 words to explain.\nOutput:"
# prompt = f"{system_prompt}\n\n**Output**\n{choices_}\n**Examples**\n{examples_}"
prompt1 = f"{system_prompt}\n\n**Examples**\n{examples_str}"
prompt2 = f"**Conversation**\n{conversation_str}\n\n**Output**\n{choices_str}"
# print(prompt1)
# print(prompt2)
messages = list()
messages.append(
{"role": "system", "content": prompt1},
)
messages.append(
{"role": "user", "content": prompt2},
)
# print(f"messages: {json.dumps(messages, ensure_ascii=False, indent=4)}")
try:
time.sleep(args.interval)
print(f"sleep: {args.interval}")
time_begin = time.time()
completion = client.chat.completions.create(
model=args.model_name,
messages=messages,
# 由于 enable_thinking 非 OpenAI 标准参数，需要通过 extra_body 传入
extra_body={"enable_thinking": False},
stream=False,
)
time_cost = time.time() - time_begin
print(f"time_cost: {time_cost}")
except Exception as e:
print(f"request failed, error type: {type(e)}, error text: {str(e)}")
continue
# print(f"completion: {completion}")
prediction_str = completion.choices[0].message.content
rid = completion.id
try:
prediction_ = json.loads(prediction_str)
prediction = prediction_["output"]
except Exception as error:
prediction_ = None
prediction = None
correct = 1 if prediction == response else 0
total += 1
total_correct += correct
score = total_correct / total
row_ = {
"idx": idx,
"rid": rid,
"messages": messages,
"response": response,
"prediction": prediction,
"prediction_": prediction_,
"correct": correct,
"total": total,
"total_correct": total_correct,
"score": score,
"time_cost": time_cost,
}
row_ = json.dumps(row_, ensure_ascii=False)
fout.write(f"{row_}\n")
fout.flush()
return
if __name__ == "__main__":
main()
|