Heh - I had no idea! Thanks for teaching me something new today
Please don’t tag me unless I’m the only person that can answer your question (which isn’t the case here).
Thanks! Would be useful to paste that into the wiki post at the top if you have a chance.
Maybe this was mentioned in a lecture and I’ve missed it, but if you want to drop into pdb automatically every time an exception occurs, execute
%pdb on (that is, you won’t need to execute
%debug to drop into the debugger).
Guess they made it up there already.
Maybe create a separate thread for the problem? I am probably not going to run the code myself today. But if this problem remains in this thread it might get missed by others given the high volume in the thread.
We have used a modified renet 34 for this task. No specific segmentation architectures are used till now. I think.
I noticed this too, and put it on my ‘check/ask it later’ list. Maybe @jeremy can chime on this one.
You could also try running docker on Windows and using Ubuntu then. However, i still don’t think it allows CUDA to be used on Windows.
Check out peterjc123 channel of PyTorch on Anaconda for Windows
soumith channel? References please.
Sure thing, sorry. My tag was mainly for the vice versa thing
Package Manager should be from Source (not conda).
This seems to be some python error on your end. The
open_image function is expecting a string as can be seen in the notebook. If you’re following the notebook, clearly you’ve missed a step somewhere.
Just take a bit of your time, and try debugging on your own. Jeremy provided a lot of tips on how to do this yesterday. Might just be the case of pulling the string out from that dict, but this is something you need to figure out on your own.
Try removing it, and let us know what you find!
No apologies required - everyone has there own preferences around at-mentions, so I’m just letting you know mine