helixfold-single issue

[jlspzw]

Install the package and requirements based on the instructions
https://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/protein_folding/helixfold-single
(BTW, is the required module cudnn: 8.10.1 version having an issue, it seems cudnn only has 8.3.x?)

when run the inference script, get the issue

Error: Can not import avx core while this file exists: /xxx/bin/python/miniconda3/envs/helixfold-single/lib/python3.7/site-packages/paddle/fluid/core_avx.so
..........
..........
from . import core_avx
ImportError: libssl.so.1.1: cannot open shared object file: No such file or directory

I checked my cpu by cat /proc/cpuinfo | grep -i avx, it seems my cpu supports avx. Could you help to check what is the problem? Thank you.

[LihangLiu]

Hi, thanks for your interest in helixfold-single. Which version of Python and PaddlePaddle are you using?

[jlspzw]

Hi, I install as the instruction, the python I just checked should be 3.7.13 and paddle should be 0.0.0. Thank you.

[LihangLiu]

This seems to be an issue related to PaddlePaddle. Could you please try to install PaddlePaddle 2.3.2 and rerun? Although Paddle 2.3.2 can't run helixfold-single, it may help us debug the problem. Thanks!

[jlspzw]

Hi, I just installed the paddle 2.3.2 with linux, conda, cuda11.2, when I tried to run python and load paddle, the issue seems to gone, but when I tried to re-run the helixfold-single, I get the issue AttributeError: module 'paddle.fluid.core_avx.ops' has no attribute 'fused_gate_attention', as you mentioned that paddle 2.3.2 may not able to run helixfold-single. So what should I do next if I want to run the helixfold-single? Thank you.

[LihangLiu]

Then it should be due to a bug in the PaddlePaddle develop version. May I ask how did you install the develop version? Are you using the wheel we provided here (https://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/protein_folding/helixfold-single#installation) or from the PaddlePaddle website or pip install?

[jlspzw]

Hi

The former version 0.0.0 I installed as the instructions as

wget https://baidu-nlp.bj.bcebos.com/PaddleHelix/HelixFold/paddlepaddle_gpu-0.0.0-cp37-cp37m-linux_x86_64.whl
python -m pip install paddlepaddle_gpu-0.0.0-cp37-cp37m-linux_x86_64.whl

the new version 2.3.2 I use the command as follows

conda install paddlepaddle-gpu==2.3.2 cudatoolkit=11.2 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/ -c conda-forge

(on this page https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/install/conda/linux-conda.html), select stable version, conda install, cuda 11.2, linux system, then I got this command. Thank you.

[LihangLiu]

Hi, thanks for your detailed information. From the feedback of PaddlePaddle team, this seems to be an incompatibility issue between the wheel package and your environment, which may related to your gcc version. Could you please try gcc --version to check your gcc version?

Besides, there are two options:

1. Wait until the end of this month. The new PaddlePaddle2.4 will be release then and should solve the problem.
2. Use the following command to compile PaddlePaddle in your machine, so that it will be compatible to your environment. (It may take 1~2 hours but may be longer depending on your machine)

https://github.com/PaddlePaddle/Paddle.git

cd Paddle

git checkout -b release/2.4 remotes/origin/release/2.4

mkdir build && cd build

cmake .. -DWITH_TESTING=OFF -DWITH_GPU=ON -DWITH_GOLANG=OFF -DWITH_STYLE_CHECK=ON -DCMAKE_INSTALL_PREFIX=$PWD/output -DWITH_DISTRIBUTE=ON -DCMAKE_BUILD_TYPE=Release -DPY_VERSION=3.7 -DPADDLE_VERSION=0.0.0

make -j40 && make install -j40 && pip uninstall paddlepaddle-gpu -y && pip install output/opt/paddle/share/wheels/paddlepaddle_gpu-0.0.0-cp37-cp37m-linux_x86_64.whl

Really sorry about the inconvenience!

[jlspzw]

Hi
My gcc version is 8.2.0, and I have tested with gcc 4.8.5 and 8.2.0 that were installed in our system, seems both not work.

I will try to build paddle from the source code, if I have an issue, I will wait for paddle 2.4 at the end of this month.

Thank you very much.

[LihangLiu]

Ok. Thanks again for your interest! We will update this issue as soon as Paddle 2.4 is released.