Making Python faster by compiling the code | Travis Oliphant and Lex Fridman
HTML-код
- Опубликовано: 25 авг 2024
- Lex Fridman Podcast full episode: • Travis Oliphant: NumPy...
Please support this podcast by checking out our sponsors:
- Novo: banknovo.com/lex
- Allform: allform.com/lex to get 20% off
- Onnit: lexfridman.com... to get up to 10% off
- Athletic Greens: athleticgreens... and use code LEX to get 1 month of fish oil
- Blinkist: blinkist.com/lex and use code LEX to get 25% off premium
GUEST BIO:
Travis Oliphant is a data scientist, entrepreneur, and creator of NumPy, SciPy, and Anaconda.
PODCAST INFO:
Podcast website: lexfridman.com...
Apple Podcasts: apple.co/2lwqZIr
Spotify: spoti.fi/2nEwCF8
RSS: lexfridman.com...
Full episodes playlist: • Lex Fridman Podcast
Clips playlist: • Lex Fridman Podcast Clips
SOCIAL:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Reddit: / lexfridman
- Support on Patreon: / lexfridman
He did a numba on them
This dude is insanely smart its kinda crazy all the great python projects he has been apart of.
Numba has helped me enormously. It’s put gpu acceleration in the hands of hobbyists. Thanks you very much.
I never understood how to utilize Numba. I've seen videos showing the speed-up but I'm having it hard to use it with various packages ..so I definately need to learn more. Or any of the JITs available in such as the one in PyTorch.
As for GPUs, have you tried CuPy ? Basically NumPy and SciPy for GPUs. You can try it out by simply importing cupy as np.
That is, write your code using numpy, for example: np.linalg.svd()
once you're done. Just change import numpy as np to import cupy as np
and everything should work. (Check compatibility with your GPU ofc)
@@Woollzable I wouldn’t say it was easy but it easier than cuda in C. I was able to work through the documentation and adapt it to my purposes.
Just use PyPy lmao
As a heavy numba user: Huge thanks for working on that and my congratulations for that huge success with such a small team! Numba is a game changer in writing fast python code!
If I said I understood 1% of that video. That would be generous to me. Still interesting.
Python packages have really stupid names.
Well, how about Julia?
I love julia it is fast syntax is really easy... If python will not become faster or something then julia is the future I think.
I know you can write python plugins with C code to make it faster, is this something different?
Yes. Now you write fast code in python and just put "@jit" in front of it.
Lucky 7
*House Numba*
~Stephen King
Micro python thoughts anyone?
I srsly hate Numba, PyPy works just fine for me tbh. Maybe one day Numba miggt actually speed up my code lol
I don't want any man I have a men that I love.
Yeet
6th!!
Why use python? There are C++ interpreters, which would permit you to execute it interactively. Then when you are ready, you compile whatever you were writing. C++ is a "real" language, you can write crazy things in C++ using objects, you can get extremely abstract and efficient.
Because C++ is supposed to be "difficult"? Is it that more difficult than correspondingly written python?
And what is the difference between what is described here and say Julia?
Julia is a chad language
Python is objectively faster to develop. On most real-world business cases, time-to-market beats time-to-run.
@@clipboarder_official Why Python is faster to develop? And which market segment has more Python software than binary software from C/C++/... if time to market is more important?
@@ciarfah Sorry, not sure what this "Chad" means in this case. Is it this one?
en.m.wikipedia.org/wiki/Chad_(slang)
@@zzip0 Python being a higher level language than c++ makes it faster to develop, think about all the features that come along with higher level languages compared to c++. Countless companies run on python based backends, data science departments often leverage python aswell. I think the overall argument is that there already are great data science tools built for python, so making it faster is an easier task than transferring all those tools to c++.
first comment again!