I’m trying to go through the provided VGG16-FCN8-CamVid Lab_1. I get an error when I try to run model.fit. See screenshot below. Any suggestions?
I’m just running the provided code. All previous cells ran without a problem, and I skipped the two workaround cells related to the data download because gdown worked.
I am also getting the same error in the COLAB provided workbook
And I have exactly the same problem with Lab 2 - the OxfordPets-UNet lab. Any suggestions?
Hi! There seems to be a bug in Colab. We pushed a fix to the notebooks in the classroom. Kindly reopen it and re-run the notebook. It’s basically adding this line up top to avoid the error.
# Install this package to use Colab's GPU for training
!apt install --allow-change-held-packages libcudnn8=8.4.1.50-1+cuda11.6
Hope it works!
Hi Chris! Thank you for the response. The update pushed correctly but I still cannot connect to GPU backend. I get the following error after running the first cell in the C2W2_Assignment.ipynb
Chris,
Thank you!
The update also works for me, so I no longer get the error when running model.fit, but I am also getting the same error message as Andre about not being able to connect to the GPU backend due to usage limits in Colab. Is this because I have been doing many GPU Colab sessions over the past 24 hours (trying to finish up the course)? Might it resolve itself this evening when demand might be lower?
I’m currently doing the lab_1 training with a CPU, but it is going VERY slowly.
Any suggestions appreciated.
Hi John and Andre! Yes Colab temporarily blocks usage after some time (you can read more about it here). It usually comes back within a day. Alternatively, if you have another Google account, you can switch to that and try to do the lab. Hope this helps!
Thanks - using another google account is a good suggestion - I should have but hadn’t thought of that. I finished the week 4 assignment without a GPU, so each epoch took about 30 minutes instead of 1 minute, and it took me a couple of tries to get the code right.
