Some questions about splitting data sets

[ywt3225]

Hello author, I followed python3.6.0 and tensorflow-1.4.0 according to your tutorial and successfully completed import tensorflow in python. There was no error in the next compilation, but when I run the split data set The same error occurred in all three lines of code.

tensorflow.python.framework.errors_impl.NotFoundError: /data/lidar/test_one/3DSSD/lib/utils/tf_ops/evaluation/tf_evaluate_so.so: undefined symbol: _ZN10tensorflow15OpKernelContext21CtxFailureWithWarningENS_6StatusE

I don't know which step of my operation my error occurred, because I strictly followed the tutorial to configure the environment, I hope you can give me some help

[Xiangzhaohong]

@ywt3225 you could try tensorflow-gpu-1.12

[amritpaul]

@Xiangzhaohong You mean we don't need to follow "pip install tensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl" and directly using "pip install tensorflow-gpu==1.12.0" ?

[Xiangzhaohong]

@amritpaul I meet some error after install tensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl , and I finally solve my error by installing tensorflow-gpu==1.12.0. I don't know why. I run this project in my docker , but I don't know how could I share my docker.

[amritpaul]

@Xiangzhaohong It will be really good if you could share your docker as it will help us to test the model. The way you can share the docker is first by saving the docker image as a tar file--> "docker save -o path_for_generated_tar_file image_name". Once you have saved it you can probably upload it on Google Drive. An alternate approach would be to share the docker build file.

[Xiangzhaohong]

@amritpaul I will try tomorrow

[amritpaul]

@amritpaul I will try tomorrow

Were you able to upload the image @Xiangzhaohong ? Even I am trying to get it working but no luck so far 👎

[Xiangzhaohong]

@amritpaul hi, sorroy for my internet. I can't upload my image to google drive, I don't know if you could use baidudisk, you can have a try. 链接: https://pan.baidu.com/s/1-O6-75v0C25BLCjo8czw1Q 提取码: iim6

[amritpaul]

@Xiangzhaohong Thanks a lot for your help. I will look into this and get back to you if needed :)

[Divadi]

Hi, regarding the error: /data/lidar/test_one/3DSSD/lib/utils/tf_ops/evaluation/tf_evaluate_so.so: undefined symbol: _ZN10tensorflow15OpKernelContext21CtxFailureWithWarningENS_6StatusE I found that replacing -D_GLIBCXX_USE_CXX11_ABI=0 with -D_GLIBCXX_USE_CXX11_ABI=1 in compile_all.sh fixes it. According to this stackoverflow post there seems to be some difference for tensorflow >= 1.15 and before.