I seem to be having an issue with the final exercise of the Deep Learning with Tensorflow 2.0 course. The instructor has about a 91% accuracy in both the vid and written down in the .ipynb supplied. However I cannot find any way to even progress above about 79-81%, even running the code supplied in the course does not yield any higher percentages. Does anyone have any idea what could cause the discrepancy?
I’ve tried using both Tensorflow 2.1.0 and Tensorflow-gpu 2.1.0, there were no differences between them.
EDIT: Even with Tensorflow 2.0.0 the course code did not provide an accuracy above 82%
Did you do the preprocessing exercise (after the preprocessing lecture in the business case)?
The 365 Team
I did. I ran the preprocessing code provided as a download, and then the solution code for the model, with the accuracy I mentioned, so about 10% below the course video’s. https://easyupload.io/qg2nei That zip contains the code I used.