Making Python faster by compiling the code | Travis Oliphant and Lex Fridman

Поділитися
Вставка
  • Опубліковано 2 гру 2024

КОМЕНТАРІ • 46

  • @Johnwilliams-th9hq
    @Johnwilliams-th9hq 3 роки тому +17

    This dude is insanely smart its kinda crazy all the great python projects he has been apart of.

  • @hansdietrich1496
    @hansdietrich1496 Рік тому

    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!

  • @phillipsmith4979
    @phillipsmith4979 3 роки тому +28

    Numba has helped me enormously. It’s put gpu acceleration in the hands of hobbyists. Thanks you very much.

    • @Woollzable
      @Woollzable 3 роки тому +2

      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)

    • @phillipsmith4979
      @phillipsmith4979 3 роки тому +2

      @@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.

    • @Agrover112
      @Agrover112 3 роки тому

      Just use PyPy lmao

  • @jb-rx3qd
    @jb-rx3qd 3 роки тому +16

    If I said I understood 1% of that video. That would be generous to me. Still interesting.

  • @kunited9
    @kunited9 3 роки тому +17

    He did a numba on them

  • @ss4yaasir524
    @ss4yaasir524 3 роки тому +1

    I know you can write python plugins with C code to make it faster, is this something different?

    • @hansdietrich1496
      @hansdietrich1496 Рік тому +2

      Yes. Now you write fast code in python and just put "@jit" in front of it.

  • @tiago08ci
    @tiago08ci 3 роки тому +5

    Well, how about Julia?

    • @xalidhemzeliyev8684
      @xalidhemzeliyev8684 3 роки тому +3

      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.

  • @kolukolev726
    @kolukolev726 3 роки тому +6

    Python packages have really stupid names.

  • @jseden1
    @jseden1 3 роки тому +2

    Lucky 7

  • @VideographerExperience
    @VideographerExperience 3 роки тому +1

    *House Numba*
    ~Stephen King

  • @michaels8297
    @michaels8297 3 роки тому

    Micro python thoughts anyone?

  • @adelinaquijano1083
    @adelinaquijano1083 Рік тому

    I don't want any man I have a men that I love.

  • @Agrover112
    @Agrover112 3 роки тому

    I srsly hate Numba, PyPy works just fine for me tbh. Maybe one day Numba miggt actually speed up my code lol

  • @zzip0
    @zzip0 3 роки тому +5

    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?

    • @ciarfah
      @ciarfah 3 роки тому +4

      Julia is a chad language

    • @clipboarder_official
      @clipboarder_official 3 роки тому +16

      Python is objectively faster to develop. On most real-world business cases, time-to-market beats time-to-run.

    • @zzip0
      @zzip0 3 роки тому

      @@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?

    • @zzip0
      @zzip0 3 роки тому

      @@ciarfah Sorry, not sure what this "Chad" means in this case. Is it this one?
      en.m.wikipedia.org/wiki/Chad_(slang)

    • @Nick-bq1ez
      @Nick-bq1ez 3 роки тому +5

      @@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++.

  • @mattboh69
    @mattboh69 3 роки тому

    Yeet

  • @frederickmusa4485
    @frederickmusa4485 3 роки тому

    6th!!

  • @Bigtuna573
    @Bigtuna573 3 роки тому

    first comment again!