Would be valuable to make a series about python accelerators, compilers, dialects or supersets. You can talk about Pyston, pyran, codon and finally mojo. And numba obviously.
My experience is that cv2 is problematic in that it is a wrapped C library. I have tried to get it to work with other Python compilers (like Numba) to no avail. (You can't even Freeze your code without a lot of special support). I use cv2 to do a lot of video enhancement (despite the fact it's not recommended by Open-CV), so I'm always looking for ways to speed the process up. Right now it takes 10-20 hours to process a movie. Would love to cut that figure down by an order of magnitude or so.
I was all excited until you said its not going to speed up numpy (so probably pandas as well). When these libraries are optimised for pypy (and other popular libraries) it will be useful. But until then it would be difficult to justify in production.
Would be valuable to make a series about python accelerators, compilers, dialects or supersets. You can talk about Pyston, pyran, codon and finally mojo. And numba obviously.
Agreed and 100% supported.
This is a really interesting topic. Thank you.
What's faster? A numpy matrix multiplier or your own pypy3 function?
Try out Codon brother. It's fire 🔥🔥
0:35 If I didn't know what you were talking about I am pretty sure your explanation would have made it more confusing..
Findhuman package useful?
How increase performance in Django web applications?
Would this do well for cv2?
With and without GPU?
I think cv2 uses numpy, so maybe it will be slower
My experience is that cv2 is problematic in that it is a wrapped C library. I have tried to get it to work with other Python compilers (like Numba) to no avail. (You can't even Freeze your code without a lot of special support). I use cv2 to do a lot of video enhancement (despite the fact it's not recommended by Open-CV), so I'm always looking for ways to speed the process up. Right now it takes 10-20 hours to process a movie. Would love to cut that figure down by an order of magnitude or so.
@@thomasgoodwin2648 Thank you for your insights.
techno tims big bro
what linux and desktop environment are you using?
@@ozu7779 Pop!_OS and i3-gaps
I was all excited until you said its not going to speed up numpy (so probably pandas as well). When these libraries are optimised for pypy (and other popular libraries) it will be useful. But until then it would be difficult to justify in production.
PyPy is nice, but sadly it's not supported by stuff like AutoPyToExe, so there is no way to produce an executable
Use numba
❤❤❤❤
Thank you. Would this make training a TF Neural Network faster?