HyRSM
HyRSM copied to clipboard
How to train on GPUs with restricted VRAM
It seems that the code cannot run on mid-end GPUs, such as NVIDIA RTX 2060s, without OOM. How can we modify the code to reduce the VRAM consumption per GPU?
It seems that the code cannot run on mid-end GPUs, such as NVIDIA RTX 2060s, without OOM. How can we modify the code to reduce the VRAM consumption per GPU?
It is recommended that you can reduce the query number (i.e., "QUERY_PER_CLASS" in yaml file) to accommodate fewer computing resources. But the performance may be different from the default setting.