Comments (6)
我也遇到了这个问题,使用vllm对第一个task测完之后就会卡住了
from opencompass.
目前可以通过运行时把 指定的很大避免对数据集进行切分从而规避这个问题,但是只是一个暂时的解决方案,还是希望开发人员可以看看怎么解决~
--max-partition-size
--max-partition-size 对单个数据集可以解决,但是多个数据集进行测评的话还是会分为多个task也会出现这个问题
from opencompass.
in opencompass/models/vllm.py
import ray
if ray.is_initialized():
self.logger.info('shutdown ray instance to avoid "Calling ray.init() again" error.')
ray.shutdown()
add above command before calling vllm LLM class;
in about 52 lines;
from opencompass.
目前可以通过运行时把 --max-partition-size
指定的很大避免对数据集进行切分从而规避这个问题,但是只是一个暂时的解决方案,还是希望开发人员可以看看怎么解决~
from opencompass.
same question, get "Calling ray.init() again after it has already been called." error
from opencompass.
in opencompass/models/vllm.py
import ray if ray.is_initialized(): self.logger.info('shutdown ray instance to avoid "Calling ray.init() again" error.') ray.shutdown()
add above command before calling vllm LLM class; in about 52 lines;
发现使用了这个方法之后,对于单模型多数据集的情况,每次有新数据集,似乎模型也要跟着重新启动一遍ray,即每处理一个数据集都会输出:
2024-04-12 01:59:04,123 INFO worker.py:1743 -- Started a local Ray instance. View the dashboard at 127.0.0.1:8266
INFO 04-12 01:59:44 llm_engine.py:75] Initializing an LLM engine (v0.4.0) with config: model='xxx', tokenizer='xxx)
(RayWorkerVllm pid=108846) INFO 04-12 02:01:37 selector.py:16] Using FlashAttention backend.
然而我发现这个过程十分耗时,有没有办法能改成启动一遍ray,一口气把数据集都跑完?
from opencompass.
Related Issues (20)
- [Bug] AttributeError: module 'pkgutil' has no attribute 'ImpImporter'. Did you mean: 'zipimporter'? HOT 1
- [Bug] Error when evaluate using LightLLM api HOT 1
- [Feature] Utilizing an Alternate Model for Model Evaluation HOT 1
- [Bug] No Pattern match or more than one config Error when running NeedleBench HOT 2
- [Bug] 为什么评估的humaneval分数比使用bigcode-evaluation-harness的分数低? HOT 5
- [Feature] Support PromptCBLUE HOT 1
- 在测评的时候显存总是有空闲,如何全部利用显存呢,单机 8*80G的 A800 HOT 6
- [Bug] alignbench 用Qwen14B infer的时候,有的prediction是空的 HOT 26
- [Bug] AlignBench Auto-j extracted judgements failed HOT 1
- [Bug] Result output is 0. HOT 7
- [Bug] open compass hangs when evaluating chat musician trained model - waiting for semaphore? HOT 11
- [Bug] Long text evaluation parameters are not clear HOT 3
- 为什么基于openai api部署的llama2-7b-chat-hf,在MMLU数据集上测试精度远低于官方数据 HOT 8
- 使用baichuan-7b评估humaneval数据与榜单差距过大 HOT 2
- [Bug] 使用NumWorkerPartitioner切分推理层和验证层任务后,最总结果指标无法将数据集结果汇总 HOT 3
- [Bug] OpenICLInfer fail HOT 16
- [Feature] 自定义数据集命令如何修改评估指标?
- [Feature] 兼容torch_npu HOT 1
- [Bug] 找不到LongContext的数据集
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from opencompass.