[Bug] 使用开源数据集sharegpt跑不起来 #219
Description
操作系统及版本
openEuler 22.03
安装工具的python环境
在anaconda/miniconda创建的python虚拟环境
python版本
3.10
AISBench工具版本
v3.1-20260319-master
AISBench执行命令
ais_bench --models vllm_api_stream_chat_multiturn --datasets sharegpt_gen --debug -m perf
模型配置文件或自定义配置文件内容
from ais_bench.benchmark.models import VLLMCustomAPIChat
from ais_bench.benchmark.utils.postprocess.model_postprocessors import extract_non_reasoning_content
models = [
dict(
attr="service",
type=VLLMCustomAPIChat,
abbr="vllm-multiturn-api-chat-stream",
path="//DeepSeek-R1_w8a8",
model="deepseek",
stream=True,
request_rate=16,
retry=2,
api_key="",
host_ip="",
host_port=8100,
url="",
max_out_len=1,
batch_size=64,
trust_remote_code=False,
generation_kwargs=dict(
temperature=0,
ignore_eos=True,
),
pred_postprocessor=dict(type=extract_non_reasoning_content),
)
]
预期行为
No response
实际行为
[2026-03-26 15:02:02,214] [ais_bench] [INFO] Warmup finished Total Count: 1 Success Count: 6 Failed Count: 0
[2026-03-26 15:02:11,329] [ais_bench] [INFO] Dataset needed memory size: 574.61133385 MB
[2026-03-26 15:02:11,329] [ais_bench] [INFO] Memory usage check passed: 20.44% < 80% (Available: 2404.41 GB)
[2026-03-26 15:02:17,272] [ais_bench] [INFO] Inference tasks completed.
/root/miniconda3/envs/ais_bench/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked shared_memory objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
[2026-03-26 15:02:17,274] [ais_bench] [INFO] Summarizing performance results...
[2026-03-26 15:02:17,274] [ais_bench] [WARNING] Can't find details perf data of [vllm-multiturn-api-chat-stream/sharegpt] in outputs/default/20260326_145848, use tmp cache data.
Traceback (most recent call last):
File "/root/miniconda3/envs/ais_bench/bin/ais_bench", line 6, in
sys.exit(main())
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/cli/main.py", line 6, in main
task_manager.run()
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/cli/task_manager.py", line 39, in run
workflow_executor.execute()
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/cli/workers.py", line 453, in execute
worker.do_work(cfg)
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/cli/workers.py", line 430, in do_work
summarizer.summarize()
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/summarizers/default_perf.py", line 427, in summarize
details_perf_datas = self._load_details_perf_data(
File "/home/zjl/scripts/bench/temp/benchmark/ais_bench/benchmark/summarizers/default_perf.py", line 251, in _load_details_perf_data
raise FileMatchError(
ais_bench.benchmark.utils.logging.exceptions.FileMatchError: [SUMM-FILE-001]can't find detail perf data file. Can't find any details perf data file in work_dir, please check outputs/default/20260326_145848. | Visit https://ais-bench-benchmark-rf.readthedocs.io/zh-cn/latest/faqs/error_codes.html#summ-file-001 for further help.
前置检查
- 我已读懂主页文档的快速入门，无法解决问题
- 我已检索过FAQ，无重复问题
- 我已搜索过现有Issue，无重复问题
- 我已更新到最新版本，问题仍存在