Hi, I am confused. Should I not be using fastai 2.0.11? I think it was recommended in the latest version of the course, but I am having a lot of issues with it. For example, tabular_learner does not use the ps argument as it used to. Is 2.0.11 a development version or how am I supposed to add dropout to the layers? Or is it generally not recommended to use the tabular_learner anymore?
1 Like
You should be. You need to pass in a tabular_config as denoted in the docs: https://docs.fast.ai/tabular.learner#tabular_learner (model parameters and Learner parameters were separated, just as they are with the language models)
For getting started with tabular I recommend reading the docs and my notebooks: https://github.com/muellerzr/Practical-Deep-Learning-for-Coders-2.0/tree/master/Tabular%20Notebooks (this was shown in the Bayesian Optimization one)
Thanks, I saw the tabluar_config before, but somehow overlooked that I can pass it to tabular_learner(). Thank you.