Excessive memory usage #2
Description
My environment is A6000, CUDA 12.9, torch=2.3.0+cu118 (CUDA is theoretically forward compatible, so there should be no major issues). The experimental parameter batch size has been reduced to 4 (which is even far below the number of classes in the source domain and target domain, so it may significantly affect clustering). The program's memory usage has reached nearly 40GB.
So how much memory is being used in your environment?