Has anyone else been experiencing extremely slow performance on the server for the past 5 days or so?
Even the import block takes up to 15 seconds. I have been working on the C2W2 assignment since Thursday, and it was running well at first but is getting progressively slower. At this point, I can’t even get through to the model.fit before one of the other blocks times out and disconnects me. I tried reverting to the unaltered assignment thinking I might be doing something that is causing this, but it is still happening. I have restarted and rebooted the kernel which sometimes gets me back in, but lately I can’t even do that.
I have even tried in the middle of the night, in case it was a traffic issue, but no dice.
Is it possible that there are some allocated resources associated only with my account that I need to release?
I resorted to running a Jupyter TensorFlow docker container locally, and downloaded the cats & dogs dataset from Kraggle, but that dataset is much smaller (1000 of each) compared to the in the assignment (~11000 of each) so even with augmentation I am unable to prevent overfitting using the techniques we have learned so far. And even if I am successful locally, I can’t get my code to run on the class server, so can’t submit.
check your codes for previous grade functions. chances are your codes are write but because of model architecture it might be burning out the kernel.
Like in train val dataset, things to check would be image size, batch size, validation split.
In create augmentation model, things to check would be input shape, and the subsequent keras layer measure for the random flip, rotation, translation and zoom.
in create model, again check your input shape and subsequent model architecture from rescaling to using conv2d, maxpooling2d, flatten and dense layers units.
model compile parameters like loss you would check the correct one if you used the correct activation in the create model as per instructions given for exercise create model.
Although if you have followed everything as per above, please also make sure you having deleted, added or edited anything outside the markers ###YOUR CODE START HERE to YOUR CODE ENDS HERE###. In case if you have done or unsure, grab a fresh copy and re-do, also I would also advise to try the next attempt tomorrow in case you have trained your model multiple times today as cpu usage limitation.
Hope this helps! let me know if you need further guidance.