> copy model to output

In the inference code:
# Copy model to output (more useful for training phase)
src="model.pth"
dest=OUT
shutil.copyfile(src, os.path.join(dest,src))

Is this part optional or required for the submission?

Posted by: rzhang229 @ Oct. 12, 2022, 8:39 p.m.

Hi,
copying the model to your output directory is optional since you're only performing inference with an already trained model. In fact, the reason why your last submission (October 12) fails is because the resulting output directory is so large the upload times out.

Posted by: kdrukker @ Oct. 12, 2022, 8:45 p.m.

Thanks! I thought that might be the problem. I will remove these lines and try again.

Best,
Ran

Posted by: rzhang229 @ Oct. 12, 2022, 8:50 p.m.

Ran,

I removed the part of the code that copies the models to our storage as we have them inside the docker image you uploaded previously. So hopefully this is smooth now. Please leave the docker images loaded in the drop-down as that is now where we are going to get your model from post challenge. You can delete anything you don't want saved but please leave the ones associated to your final submission(s).

Posted by: challenge-organizer @ Oct. 13, 2022, 8:53 p.m.
Post in this thread