I think it's a great tool to use and not use python because it's so easy to use If you save the visualization result as it is in the script or start it on the workstation, it can be accessed from the outside, and it is very useful for research. However, I was a little stuck with the Jupyter notebook.
It's multiprocessing. If you use it normally, there is no problem at all, but if you occasionally throw an error and stop, the process will not stop unless you restart the notebook ... I finally found out after looking for a reason.
I think it's not uncommon to write a script to some extent and then perform various processing after executing % run hogehoge.py
.
It seems that there is a restriction that the parallelization target must be the function that can be called by the first % run hogehoge.py
or the function written directly in jupyter.
In other words, when % run hogehoge.py
followed by % run fugafuga.py
but the function of fugafuga.py
cannot be called by hoghoge.py
is parallelized. It seems that I get an AttributeError (main gets angry when I get angry, and freezes).
There is no problem for people who do everything with Jupyter scripts, but when it comes to big projects, you need to be careful when trying various different scripts instead of test code. (Can you convey this explanation ...)
Recommended Posts