search

mahmoodlab / CLAM

Data-efficient and weakly supervised computational pathology on whole slide images - Nature Biomedical Engineering
http://clam.mahmoodlab.org
GNU General Public License v3.0
1.02k stars 340 forks source link

I am experiencing 0% GPU utilization when using the NVIDIA RTX 3090 GPU for feature extraction. #196

Closed Liuyifan1016 closed 5 months ago

Liuyifan1016 commented 1 year ago

When using extract_features_fp, I'm experiencing significantly slow performance, and upon checking the GPU utilization with nvidia-smi, it shows 0% usage. How can I resolve this issue?

scjjb commented 1 year ago

@Liuyifan1016 It's likely that you have other resource constraints meaning that your GPU can barely get going. There is a huge amount of data being loaded, and if your input/output speed isn't very very fast then your GPU will just be sat waiting.

Liuyifan1016 commented 1 year ago

@Liuyifan1016 It's likely that you have other resource constraints meaning that your GPU can barely get going. There is a huge amount of data being loaded, and if your input/output speed isn't very very fast then your GPU will just be sat waiting.

Thank you for your response. I have resolved the issue. GPUs of the 30 series and above require CUDA 11.0 or higher. The problem was caused by installing a lower version of CUDA as specified in the configuration file, which resulted in the inability to utilize the GPU. The solution was to update CUDA to a version compatible with the 30 series GPUs or use a GPU series that supports CUDA 10.