I have completed Ex_3 fine-tuning, but the kernel has died several times around epoch 9 while running inference. This is slowing things down and the kernel may continue to die at this point until the end of time. Will the autograder penalise me if I submit the assignment with less than 10 epochs completely run?
There is no known reason for the kernel to die at a particular iteration.
The auto-grader doesn’t use your kernel. It runs your notebook in a separate environment.
Thank you TMosh, I have submitted (and scored 100%).
One potential cause of this sort of issue is putting voluminous print statements in your inner loops. Every output increases the memory footprint of the notebook in memory and that can eventually cause things to die.
Or it could be that you just got unlucky and encountered “bad weather” in the AWS Cloud where the VM you were executing on was very busy with other multitasking operations.