Environment requirements

[zijianh4]

Hi, I wanna reproduce your work and I am trying to set up the environment. Although I use the requirements.txt to install, it seems that there is an error that says "AttributeError: 'GATConv' object has no attribute 'lin_src'". Currently, I am using python==3.7.2, torch==1.10.1+cu113, torch-geometric==2.0.3 ,torchvision==0.11.2+cu113, tensorboardX==2.5 and seaborn==0.10.1. Could you please tell me the environment that you set up? Thx!

[AlexDuvalinho]

Hi,

Yes it should be due to your environment, no one has reported this pb before. I just tried to run all scripts with python==3.7.10, pytorch==1.8.1, torch-cluster==1.5.9, torch-geometric==1.7.0, torch-scatter==2.0.6, torch-sparse==0.6.9, seaborn==0.11.1, tensorboard==1.15.0; and it worked…

I don’t recall you need very strict version of packages to run that code so you might have been unlucky here. Either try the ones above or simply replace >= with == in the requirements.txt and it should work just fine. If it still does not, you can play with some packages configurations and it shall be all right.

I am not sure I will be of more help on that — but if you have other questions I will be happy to help.

Best, Alexandre

----- Mail original ----- De: "Zijian Huang" @.> À: "AlexDuvalinho/GraphSVX" @.> Cc: "Subscribed" @.***> Envoyé: Vendredi 4 Mars 2022 03:57:56 Objet: [AlexDuvalinho/GraphSVX] Environment requirements (Issue #2)

Hi, I wanna reproduce your work and I am trying to set up the environment. Although I use the requirements.txt to install, it seems that there is an error that says " AttributeError: 'GATConv' object has no attribute 'lin_src' ". Currently, I am using python==3.7.2, torch==1.10.1+cu113, torch-geometric==2.0.3 ,torchvision==0.11.2+cu113, tensorboardX==2.5 and seaborn==0.10.1. Could you please tell me the environment that you set up? Thx!

— Reply to this email directly, view it on GitHub , or unsubscribe . Triage notifications on the go with GitHub Mobile for iOS or Android . You are receiving this because you are subscribed to this thread. Message ID: <AlexDuvalinho/GraphSVX/issues/2 @ github . com>

[zijianh4]

Hi, Yes it should be due to your environment, no one has reported this pb before. I just tried to run all scripts with python==3.7.10, pytorch==1.8.1, torch-cluster==1.5.9, torch-geometric==1.7.0, torch-scatter==2.0.6, torch-sparse==0.6.9, seaborn==0.11.1, tensorboard==1.15.0; and it worked… I don’t recall you need very strict version of packages to run that code so you might have been unlucky here. Either try the ones above or simply replace >= with == in the requirements.txt and it should work just fine. If it still does not, you can play with some packages configurations and it shall be all right. I am not sure I will be of more help on that — but if you have other questions I will be happy to help. Best, Alexandre …

Hi, I just wonder if the environment plays with CUDA or not. If I use cuda, should I follow the requirements for Colab specific in requirements.txt? Thx!

Zijian

[AlexDuvalinho]

If you use cuda, you should definitely take it into consideration when installing the packages. Depending on your version of cuda, you can use the commands specified in the requirements.txt (commented out). Otherwise, I invite you to look at pytorch / pytorch-geometric guides to install cuda compatible packages.

----- Mail original ----- De: "Zijian Huang" @.> À: "AlexDuvalinho/GraphSVX" @.> Cc: "AlexDuvalinho" @.>, "Comment" @.> Envoyé: Vendredi 4 Mars 2022 13:18:16 Objet: Re: [AlexDuvalinho/GraphSVX] Environment requirements (Issue #2)

Hi, Yes it should be due to your environment, no one has reported this pb before. I just tried to run all scripts with python==3.7.10, pytorch==1.8.1, torch-cluster==1.5.9, torch-geometric==1.7.0, torch-scatter==2.0.6, torch-sparse==0.6.9, seaborn==0.11.1, tensorboard==1.15.0; and it worked… I don’t recall you need very strict version of packages to run that code so you might have been unlucky here. Either try the ones above or simply replace >= with == in the requirements.txt and it should work just fine. If it still does not, you can play with some packages configurations and it shall be all right. I am not sure I will be of more help on that — but if you have other questions I will be happy to help. Best, Alexandre … Hi, I just wonder if the environment plays with CUDA or not. If I use cuda, should I follow the requirements for Colab specific in requirements.txt? Thx!

Zijian

— Reply to this email directly, view it on GitHub , or unsubscribe . Triage notifications on the go with GitHub Mobile for iOS or Android . You are receiving this because you commented. Message ID: <AlexDuvalinho/GraphSVX/issues/2/1059406305 @ github . com>

[zijianh4]

If you use cuda, you should definitely take it into consideration when installing the packages. Depending on your version of cuda, you can use the commands specified in the requirements.txt (commented out). Otherwise, I invite you to look at pytorch / pytorch-geometric guides to install cuda compatible packages. …

Hi, Thx for your reply. However, I think there may be some inconsistency in the libraries. Since when I try to use torch-geometric==2.0.3, there is the previous bug AttributeError: 'GATConv' object has no attribute 'lin_src', which I find it is because in version 2.0.3, there is lin_src and lin_out instead of l_src and l_out in version 1.7.0. However, when I try to use version 1.7.0, there is another error ModuleNotFoundError: No module named 'torch_geometric.data.storage'. This is because there is such a module storage.py in version 2.0.3 but not for version 1.7.0. Is this because the code or the model is not updated? Thx!

[Yeo-Jun-Choi]

I recommend you refer to this https://github.com/pyg-team/pytorch_geometric/issues/3868