search

XPixelGroup / BasicSR

Open Source Image and Video Restoration Toolbox for Super-resolution, Denoise, Deblurring, etc. Currently, it includes EDSR, RCAN, SRResNet, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, ECBSR, etc. Also support StyleGAN2, DFDNet.
https://basicsr.readthedocs.io/en/latest/
Apache License 2.0
6.66k stars 1.17k forks source link

out of memory on testing #350

Open joneschunghk opened 3 years ago

joneschunghk commented 3 years ago

I set the batch size to 4 during training and it works fine. However, when I tried to test, it resulted in "out of memory" CUDA errors. And I can't find any settings to reduce the amount of VRAM usage. Will it be effective to modify "num_feat" and "num_block"

hhwucn commented 3 years ago

You probably don't have enough video memory, I had the same problem when I was testing the EDSR network on Div2K data sets (I was using 10G video memory).Both the Torch environment and the model run itself take up a lot of video memory, which can happen if you load too much data.

tongchangD commented 3 years ago

how to slove it