It would be great to have more bearings & context for how the different fastai classes fit into the larger ecosystem of tools available for machine learning & deep learning.
Is the following correct?
p1 v1: tensorflow / theano / keras / python2
p2 v1: pytorch / python3
p1 v2: fastai / pytorch / python3
p2 v2: fastai / pytorch / python3
Also verrrrrrry interesting: how to find this out?
Is reading the “includes” in a notebook (and in the included libraries such as fastai) the best way? Do people have quick commands from keras or pytorch that they run to verify for themselves what’s going on?
All correct! If you want to dig deeper, you may:
- Interactively, use the notebook’s
?? operator, as well as Shift+Tab in each Jupyter Notebook’s cell.
- Read the imports and follow the trend.
To learn more about the function
ConvLearner.pretrained, do this in a blank cell of your notebook:
That’ll get you a view of the function’s docstring, and
that’ll get you the source code of the function.
You’re on a great path, so please ask more questions! Glad to help
In short, fast.ai is adapting itself to the latest techs…
Is it correct though? I’m getting less sure that p1v1 used tf.
I can’t see any notebook mentioning tf modules either. I must have been mistaking p1v1 and p2v1.
A lot of modules are being imported automatically when you import fast.ai