got nan during sampling

[cuicanyu]

It's a good job, but I got pre_lig_pos NAN when I used the pretrained ckpt to sampling ligands. Could you please check the sampling code?

[DongkiKim95]

Thanks for your insightful work and for providing code implementation.
Unfortunately, I am in the same situation where the values of sampled positions explode and then become NAN.

Is there anyone who solved this problem?

[FeilongWuHaa]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

[cuicanyu]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.
I used this:
python sample_split.py --start_index 0 --end_index 99 --batch_size 25

[cuicanyu]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply.

[Kartinaa]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply.

Have you solved this problem? I also get nan for my generation using sample_split.py

[YangLing0818]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply.

Thanks for your interests in our work. For your problem:

1. Check if the paths in the training.yaml and sampling.yaml files within the config directory are correctly configured, including the dataset and pretrained models.
2. Verify that the packages in the environment, especially the versions of pyg and torch-geometric, are right.
3. You can also refer to this issue Request for Model Checkpoints or configuration #2 (comment)

[Kartinaa]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply.

Thanks for your interests in our work. For your problem:

1. Check if the paths in the training.yaml and sampling.yaml files within the config directory are correctly configured, including the dataset and pretrained models.
2. Verify that the packages in the environment, especially the versions of pyg and torch-geometric, are right.
3. You can also refer to this issue Request for Model Checkpoints or configuration #2 (comment)

Thanks for your reply! Can I ask how much GPU memory you were using for sampling the ligands? I encountered RuntimeError: CUDA out of memory on my RTX 4070s even with 1 as batch size, I'll look for better GPU resources but just curious about what GPU you are using for batch_size = 25.

[cuicanyu]

Thanks for your reply. I have solved the problem by rebuilding the virtual environment.   ***@***.***  

…

------------------ 原始邮件 ------------------ 发件人: "YangLing0818/IPDiff" ***@***.***>; 发送时间: 2024年10月14日(星期一) 上午9:24 ***@***.***>; 抄送: " ***@***.******@***.***>; 主题: Re: [YangLing0818/IPDiff] got nan during sampling (Issue #4) which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working. Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply. Thanks for your interests in our work. For your problem: Check if the paths in the training.yaml and sampling.yaml files within the config directory are correctly configured, including the dataset and pretrained models. Verify that the packages in the environment, especially the versions of pyg and torch-geometric, are right. You can also refer to this issue Request for Model Checkpoints or configuration #2 (comment) Thanks for your reply! Can I ask how much GPU memory you were using for sampling the ligands? I encountered RuntimeError: CUDA out of memory on my RTX 4070s even with 1 as batch size, I'll look for better GPU resources but just curious about what GPU you are using for batch_size = 25. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[JazZyzJ]

which py file or command did you use? the scripts/sample_for_pocket.py has some error , not working.

Excuse me, can you use this code to generate normal molecules? I am currently conducting research in this area, but according to the author's read me and using the author's pre trained model, I have been encountering situations where molecular coordinates show NAN during the sampling process. May I ask how you sampled it? We look forward to your reply.

Thanks for your interests in our work. For your problem:

1. Check if the paths in the training.yaml and sampling.yaml files within the config directory are correctly configured, including the dataset and pretrained models.
2. Verify that the packages in the environment, especially the versions of pyg and torch-geometric, are right.
3. You can also refer to this issue Request for Model Checkpoints or configuration #2 (comment)

Thanks for your reply! Can I ask how much GPU memory you were using for sampling the ligands? I encountered RuntimeError: CUDA out of memory on my RTX 4070s even with 1 as batch size, I'll look for better GPU resources but just curious about what GPU you are using for batch_size = 25.

The same. Have you solved it?