cupy.cuda.memory.OutOfMemoryError: out of memory to allocate

There is an error as per the title.

 python gpucuda cudnn chain

Please tell me the countermeasures.
Due to the program configuration, I run del and process images, but I can't reduce the image size, reduce the batch size, or change the network...

I'm sorry for being selfish, but I appreciate your cooperation.

python chainer cuda gpu

2022-09-30 20:21

1 Answers

If you don't want to use multiple GPUs, there is a function called F.forget. This function can free memory by removing the intermediate feature .array.

Also, if you can use multi-GPU, Data parallelization using multi-GPU is also helpful.
In this case, a minimum batch size of 1 must pass forward-backward.

2022-09-30 20:21

If you have any answers or tips

Popular Tags

python x 4648

android x 1593

java x 1494

javascript x 1427

c x 927

c++ x 878

ruby-on-rails x 696

php x 692

python3 x 685

html x 656