CUDA out of memory

[DawnJYe]

Does running this model require a lot of GPU memory? Will using an A100 GPU still result in an out-of-memory error?

[DawnJYe]

Hello, we tried to solve the issue.

This is what we did:

Modify the Graphormer3D model to use gradient checkpointing and reduce batch size to decrease GPU memory usage.

You can review changes in this commit: trinks-slam818@15984ac.

Caution

Disclaimer: The concept of solution was created by AI and you should never copy paste this code before you check the correctness of generated code. Solution might not be complete, you should use this code as an inspiration only.

Latta AI seeks to solve problems in open source projects as part of its mission to support developers around the world. Learn more about our mission at https://latta.ai/ourmission . If you no longer want Latta AI to attempt solving issues on your repository, you can block this account.

I don't use graphormer3D and I just use base graphormer.It seems that the batchsize needs to be reduced to very small for normal training.