Shulou(Shulou.com)06/01 Report--

How to understand the visual memory mechanism of pytorch torch.cuda.empty_cache (), in view of this problem, this article introduces the corresponding analysis and solution in detail, hoping to help more partners who want to solve this problem to find a more simple and feasible method.

Pytorch has been able to automatically reclaim video memory that we do not use, similar to python's reference mechanism, when the data in a certain memory no longer has any variable references, this part of the memory will be freed. However, it is important to note that when some of the video memory is no longer in use, the released video memory is not visible through the Nvidia-smi command, for example:

Device = torch.device ('cuda:0') # defines two tensordummy_tensor_4 = torch.randn (120,3,512,512). Float (). To (device) # 120 "3" 512 "512" 412 "4000max 1000 = 377.48Mdummy_tensor_5 = torch.randn (80,3,512) 512) .float () .to (device) # 80'3' 512 '512' 4bank 1000 251.64M# and then release dummy_tensor_4 = dummy_tensor_4.cpu () dummy_tensor_2 = dummy_tensor_2.cpu () # here although the video memory above has been released But we can see through the Nvidia-smi command that the video memory is still occupying torch.cuda.empty_cache () # only after executing the above sentence will the video memory be released in the Nvidia-smi

The developers of Pytorch also explained that this part of the released video memory is available, but it is not displayed in Nvidia-smi.

This is the answer to the question about how to understand pytorch's video memory mechanism torch.cuda.empty_cache (). I hope the above content can be of some help to you. If you still have a lot of doubts to be solved, you can follow the industry information channel for more related knowledge.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.