my model runs - but really slowly im not sure if ill get to 95 average structural similarity index but if its super slow there must be something im not doing right it needs more tuning i suppose - no error messages but model is finicky - if i change loss function i draw a trace error - can i get advice on what i might try - am i allowed to make the classifier more complex and deeper instruction implies i have to match expected output -(i’m guessing i cant just grab the inception model using transfer learning) can i try data augmentation using tensorflow datagen ?can i play with the optimizer in one spot criteria seems to state i must use rms prop with a learning rate .01 - and when we move to 18 epochs it says to use exact same parameters except prefix -i trimmed some fat and now the model is moving- one other point in the instructions at the def augment it asks us to resize the model and if im not mistaken in the do_salience comments its asks us to do it again isnt that redundant and wont it slow the model also ? i topping out at 87 like some others i read so im thinking its one line of code here
1 Like
Hi Paul. Sorry for the late reply. There’s a recent issue with Colab that disables the use of GPUs for some labs in this course. Learners who reported this lab got around it by using the fallback runtime. Hope this helps.
1 Like
Hello @chris.favila
Chris the learner had created multiple posts for the same assignment. His issue is resolved now.
Regards
DP
2 Likes
Great! Thanks Deepti!
1 Like