201723201401012

Results 10 comments of 201723201401012

> 导入其他的pt模型文件报错: PytorchStreamReader failed locating file constants.pkl: file not found 请问是什么问题? 我也是这问题,是不支持么?

PyTorch引擎会回收

> @201723201401012 I ran your code and I didn't see any error, no OOM, the max memory usage is about 6G on my machine. 我测试也是维持6G左右,但是这个6G并不会回收,你连续压测内存还会上涨,直到oom,你可以连续并发100次压测,你将会看到问题

> > @201723201401012 I ran your code and I didn't see any error, no OOM, the max memory usage is about 6G on my machine. > > 请问,能不能回复一下我的问题(就在当前这位作者的下面就是我的提问),目前代码运行在CentOS8服务器上,我使用接口请求工具进行测试了,如果不是多线程并发请求,内存波动基本平稳不是特别大,但是只要是并发请求,假设10个,内存就会有数十GB的增长,随着并发请求的次数增多内存也是一直在增长,如果只是一瞬间有特别大的内存占用也没什么问题,服务器内存空间完全够用,但是问题就是并发请求完了内存不释放,有个两三次并发请求,内存占用就能超过100GB。我想知道paddleOCR是我哪里使用的不对吗,你们官方在开发测试过程中使用的都是什么环境,比如:是不是基于Spring?JDK版本?打包和部署方式?还是我的代码有问题?谢谢! 你的问题基本和我一样,并不会回收,我这个问题如果用PyTorch引擎内存会回收,但是pytorch引擎在aarch64下性能非常低,有十倍的差距

> 1. I tested your code with 300,00 iteration (takes 20 minutes), no memory leak, > 2. The memory consumption is related to number threads, in your test you are...

> what error do you see? When you say OOM, it java heap memory error or native memory error? java

@frankfliu python可以正常运行,我的基础镜像用的是ubuntu20,里面就下了jdk和tomcat,我们有两份环境x86和aarch64 ,aarch64推理就很慢,不知道如何排查,https://github.com/deepjavalibrary/djl/blob/master/.github/workflows/native_jni_s3_pytorch.yml#L264-L273 这个里面用了amazonlinux:2这个镜像,难道我要切换到这个基础镜像?

@frankfliu Can you help me? This is very important to me

> Most likely is related to preprocessing. You can compare the forward only pass between python and java, they should be identical. 你可以看下我的代码么

> Most likely is related to preprocessing. You can compare the forward only pass between python and java, they should be identical. 在中国有群么,我看b站上那个群的二维码已经失效了