Instead of running several notebooks on the same GPU, is it possible to run a single notebook and have cells run in parallel with each other? For quickly testing things out it would be convenient to train multiple models at the same time on the same notebook for the sake of keeping everything in one place. I was curious if anyone else has done something like this or if this is even possible. I haven’t seen anything on the Fastai forums…
On a similar vein, can we get a notebook’s cell to process and return async? Maybe via some jupyter-magic command? E.g. While cell-1 is running for say ten minutes, could I run a simple command in cell-2 and return the result before cell-1 is finished executing in the example below:
In [*] : learn.fit(10) # takes ten minutes
In [2] : print(type(learn))
output2 >‘fastai.Learner’
Obviously you can open a new notebook for quick “one-off” stuff, but you can’t drill into the existing variables within your notebook’s current namespace/state with a fresh notebook.
Anyone maybe know another data science project where they can do this?