Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

DGL 0.7.2 installed via conda seems to link an non-existent version of pytorch 1.10.2 libary

See original GitHub issue

🐛 Bug

It seems that dgl 0.7.2 supports up to pytorch 1.10.0. package dependency logic links it to /lib/python3.9/site-packages/dgl/tensoradapter/pytorch/libtensoradapter_pytorch_1.10.2.so, but under this path I can only find libtensoradapter_pytorch_1.10.0.so

To Reproduce

Steps to reproduce the behavior: I use conda 4.11.0, python 3.9, all other packages installed using conda-forge. Using brand-new conda environment. Pytorch installed here is 1.10.2, dgl is 0.7.2

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
conda install -c dglteam dgl-cuda11.3
go to python and import torch, import dgl

Error Message: TensorDispatcher: dlopen failed: {my environment path}/lib/python3.9/site-packages/dgl/tensoradapter/pytorch/libtensoradapter_pytorch_1.10.2.so: cannot open shared object file: No such file or directory

Environment

DGL Version: 0.7.2
Backend Library & Version: PyTorch 1.10.2
OS : Ubuntu 20.04LTS
How you installed DGL (conda, pip, source): conda
Build command you used (if compiling from source): N/A
Python version: 3.9.10
CUDA/cuDNN version (if applicable): 11.3/8.2.0
GPU models and configuration (e.g. V100): RTX A5000
Any other relevant information:

Additional context

Issue Analytics

State:
Created 2 years ago
Reactions:1
Comments:6

Top GitHub Comments

1reaction

Rhett-Yingcommented, Feb 18, 2022

It’s ok to ignore this error message as this lib is for better performance(memory efficiency) only.

I am also seeing this message using DGL from conda. Would that be possible to silent it? It tends to pollute my logs which make it hard for debugging purposes.

we’re working on a patch to suppress it. will ping you once ready.

1reaction

luobroadcommented, Feb 17, 2022

If you really want you can downgrade your PyTorch to 1.10.0, what I am doing right now.

On Thu, Feb 17, 2022 at 15:32 Hadrien Mary @.***> wrote:

It’s ok to ignore this error message as this lib is for better performance(memory efficiency) only.

I am also seeing this message using DGL from conda. Would that be possible to silent it? It tends to pollute my logs which make it hard for debugging purposes.

— Reply to this email directly, view it on GitHub https://github.com/dmlc/dgl/issues/3716#issuecomment-1043400466, or unsubscribe https://github.com/notifications/unsubscribe-auth/AW3G3LNDZ4IJAYBVMRFWWITU3VLNFANCNFSM5NWIBKKA . You are receiving this because you authored the thread.Message ID: @.***>