kunjiefan / MLEC-iSL

A graph transformer model for cell-specific synthetic lethality prediction
MIT License
0 stars 0 forks source link

CUDA out of memory #1

Open mrxgavin opened 1 year ago

mrxgavin commented 1 year ago

Hi, may I ask how much GPU memory is needed to train the model? Is there any way to modify the batch size?

kunjiefan commented 1 year ago

Hi, Thanks for your interest.

I can manage to run the model using an Nvidia V100 with 32 GB configuration.

I don't use batch training in this model. If you want to run the model with less CUDA memory, you can try to modify hyper-parameters associated with the graph transformer model, such as "head" and "embed_size".