The performance of my model was awful. I noticed that after the preprocess function,
We do not have an “Optimize the dataset for training” cell like we had in Lab 1 of week 3. with that, we do not cache(), shuffle() and prefetch(). Could this be the reason for the bad performance?
hi @eciuffo
did you recall the input and weights correctly?
Thank you so much Deepti. It turned out that my weight was still set to None. After some tweaking on the learning rate, I achieved a 100% accuracy.
1 Like
I’m having issues with this. If I change the weights from “None” to LOCAL_WEIGHTS_FILE it doesn’t do anything. When I first tried to add separate code to load the weights file, it improved and hit the target by epoch 33 but I’m not seeing what is described here.
I have been playing with it and now when I add the separate code for loading weights file, it doesn’t do anything anymore
Joshua try to create a new topic and explain the issue you experiencing.
mentioning that you mention the weight, will not make mentors understand what is the issue you are encountering.
if it is about not able to get the desired accuracy, then the hint would be keep your model as simple as possible, also refer the videos and ungraded lab which provides you more hint on how simple the model needs to be.
Until you explain what was issue, I can only guide this much.
Regards
DP
Thanks Deepti. Actually seems to be working now. I started over and ran each cell from the beginning. Something must have been confused with the repeated changes.
1 Like