deep-learning-with-python-notebooks
deep-learning-with-python-notebooks copied to clipboard
fchollet
Chap. 6.3, recurrent dropout, cannot reproduce book's results with TF2.2
Keras: 2.4.3 Tensorflow: 2.2
Running parts 6.3.6 (recurrent dropout) and 6.3.7 (stacked recurrent layers) of the original notebook gives the following training and validation losses:
6.3.6:
6.3.7:
How do we have to modify the hyperparameters to reproduce the book's results?
Attached is my environment.yml if someone wants to reproduce my findings.
