search

mpriessner / CAFI

Combined Repos from DAIN and Zooming SlowMo
MIT License
19 stars 5 forks source link

RuntimeError: cuda runtime error (77) #2

Open SujanGhim opened 2 years ago

SujanGhim commented 2 years ago

Hi, I tried to run the ZoomingSlowMo_4_Microscopy.ipynb notebook with mine own and also provided datasets, however it throwed following error in section 4.3.1 with both datasets: RuntimeError: cuda runtime error (77) : an illegal memory access was encountered at /pytorch/aten/src/THC/THCCachingHostAllocator.cpp:296

In another case, I was trying to use a 5D dataset (TZXYC). When I give the file in section 3 it gives following result: "The TIF image dimensions are (20, 19, 2, 1024, 1024)". I guess it was supposed to give the image dimensions as (20, 19, 1024, 1024, 2). Probably because of this it couldn't prepare the data for training. Then I splitted channels and reached until section 4.3.1. Please let me know if you need further information. I am interested to use the notebook for Airyscan live cell imaging data which are of between 20-50 timepoints, around 20 z-slices and less than 1024x1024 in dimensions.

I appreciate your quick response.

Regards, Sujan

mpriessner commented 2 years ago

Hello Sujan, Generally I would suggest to work on files that have just one channel. The code then artifically creates 3 channels to match the input dimensions of the pretrained network. If you just provide 2 channels this might be the reason why it comes to some problems. Also try to provide it with 512x512 sized data. Maybe this is the reason for that problem you are encountering. Let me know if you need further help.