what a fricking surprise, the 7 month old machine learning model i need to build a thing doesn't fricking work because it expects a fricktarded version of numpy
time for venv #2385627835235
oh what's that??? they didn't include their moronic choice in requirements.txt so now you have to also larp as a moron to use this stupid piece of shit?
"we are adults" say the ML people
 DMT Has Friends For Me Shirt $21.68 |
 |
DMT Has Friends For Me Shirt $21.68 |
> too moronic for python
> literally filtered
it's not challenging
it's not difficult
it's not even that tedious
it's just a fricking abortion compared to what computer programming languages should be
absolutely filtered keeeeeeeeeeeeeeeek
It's the same kind of midwit that'd shame musk for asking for help with setting up python. IQ is a function of problem solving efficiency, or quantity over time. With enough effort you could hope to overcome "talent", ie. IQ. But fact is a high IQ individual will run circles around you overall for the same amount of effort. If you spent the same amount of effort messing with python deps, you'd have a more comprehensive handle on it than the midwit python simp, and quicker at that too.
>i've been trapped in kindergarten with them long enough to be elitist about it
there is a reason i am a real legal engineer and you are not my friend
who are you quoting?
>a real legal engineer
a python dependency engineer?
scurry on lad, you're not legally allowed to sign off on things because there is simply no jurisdiction which considers programmers to be actual engineers and professionals.
it's why dijkstra did what he did, he realized he was inadvertently founding a field of dunning krugerite manchildren who he despised.
I don't disagree, but if you're an engineer and you have the need to use any non-vendor specific software but a CAD (and only to design) you're not a real engineer, you're a cog in the wheel working under king grifter ie. your boss. there's a reason why technicians and those on the ground that actually work look down on you "people"
t. helped out at a business designing and building single order precision machines for automation at large multinationals that regularly saw basedgineers get the frick out when they find out what real engineering work entails
you make me feel something
You don't get it man. Python dependency management is hell.
p o e t r y
o
e
t
r
y
I tried to install some packages lately. Say, it has >=2.2 in requirements, 2.3 is installed. Pip starts pulling 2.6, breaking every shit and its DB in the process. Thanks, pipo. Meanwhile, Poetry builds its own dependency graph and is just works, like an OS package manager.
>poetry
Is it any better than venv?
Just use pip-tools with the hashes option. poetry is bloat, pulls in like 50 dependencies.
despite that, i think it does a better job than venv/conda
Just build a clean environment for each project and keep everything sandboxed
I confirm, I built caffe and pytorch with rocm support. MLers like to larp as some geniuses but in reality they can't handle build scripts. Can't wait till programmers get replaced with AI and everyone switches to ML, they'll see how competent people work. Especially waiting for their reaction when they realize that absolutely everyone learnt probability theory and statistics in uni and the only reason nobody worked in ML was that nobody was interested enough.
The problem isn't Python or even pip the problem is dickheads and idiots writing unstable library APIs and pathetic babies who cry about using virtualenv.
The stupidest people of all are javascript zoomers who think that the problem with careless, unstable library APIs is that there isn't a bloated, overly complex monstrosity of a build tool to sweep all that rotting garbage under the rug.
>pykiddies on edge since they got exposed for murdering 4 people on a sub
this is what happens when people take a toy language seriously
on edge since they got exposed for murdering 4 people on a sub
>this is what happens when people take a toy language seriously
Python is a overglorified TI-84, now fix your requirements.txt, cuckdev.
Is there anyone working on a high(er) level language that targets machine learning?
Skill issue
python dependency management was literally made by complete delusional morons which is in line with the demographic
its a really good language, it's just that the MLtards embarrass it.
at least with a real language like C you get true errors not just "we are a schizophrenic community with no cohesion or leadership"
Just change the requirements.txt it probably isn’t as version dependent as fhat
requirements.txt was missing like two things, i expected that though
use anconda anon
https://www.anaconda.com/
it was data scientist use, try and find a docker container too if you can. I used to struggle with this too til I put in effort understanding this.
>Docker
Why is this important
t. I'm a moron
containers, so in Op example, he has a month old project that doesn't work because of dependencies failure i.e. incompatible version. If the original authors provide a docker file which they often do for data science/ML projects, it freezes the project at that time so that you can run it later without worrying about user configuration errors i.e. requirement.txt not doing enough to keep track of things going well. Might be a hassle to setup, ask GPT4 to spoonfeed using docker
they provided a requirements.txt but it was woefully insufficient
also it was not a pajeet, this was a weeb, so probably a sudaca, or a chicano in america. only brown non-pajeets like anime.
the most damning fact about the whole ML-python community is the fact that they spam emojis like they're fricking toddlers
I definitely wont' bother with anything made by a weeb or without at least one white, slavic or jap name on the team after this though.
i thought since i was trying to identify anime to replace it the weebs would have the best anime detector model. they do not, they are incompetent brown third worlder wogs as always.
requirements.txt do not work
Or at least they don't as a long term solution. The problem is, most projects only keep a subset of versions on Pypi. So in 6 months that version of shitpyckage you put in your requirements.txt might just not be there
>requirements.txt do not work
>Or at least they don't as a long term solution
what a fricking joke
You write them incorrectly if they don't work for you. Like all dependency management, the publisher is solely to blame (or praise) for a well written requirements file.
Generally, a `pip freeze > requirements.txt` completely solves the issue and creates a fixed recipe for the environment. Where things go wrong is publishers thinking they know better and altering requirements.txt manually.
Not a Python problem, this applies to every package or program that has dependencies of any kind.
>Generally, a `pip freeze > requirements.txt` completely solves the issue
No it doesn't. This is exactly what I'm talking about. The pip freeze will only work for someone who uses the same system, and the same system right now. In 6 months half those exact version numbers won't be available. And if your recipient is on Mac/windows, he will absolutely never have access to the exact same package versions.
yeah in my case they had no spec on numpy and it didn't like me 1.2.4 or some shit so i had to take the venv to 1.1.9
at that point it would have worked, but i didn't want to fix the symlink on my cuda driver thing, i'll just install 12 soon anyway
How is "packages not being available" a problem with requirements, or even Python? Yes, someone removing a dependency of your project is a problem for YOU, and it is YOUR decision to pass this problem to your users.
The correct thing to do is to bundle the dependencies with your program or to put them on YOUR infrastructure.
requirements.txt fully supports local packages, as well as custom URLs to private repos.
Again, it is your decision to insert this problem into your program, and this is not limited to Python in any way. If you told users that XP is the only supported OS or CUDA 3.0 is the only supportrd runtime then you'd be creating the same exact problem for them.
Not a problem on grown up package managers.
Never had this kind of issue with maven, cargo, nix, etc.
>t. clueless
It's not a problem with pip either, OP is brain dead. I never saw a package going missing from PyPi and I deal with these as part of my job.
>and I deal with these as part of my job.
Do you maintain some packages yourself?
Yes, internal to our company. We use everything from webshit to pytorch to boto3 to Selenium.
take a look mate
doesn't work with regular debian install which is 1.24.3, needs 1.19
https://github.com/cheese-roll/light-anime-face-detector
(i was looking at using it to detect anime so i can turn it into non anime and wanted something that gave results fast.)
your entire field and all of the people who you interact with are extremely childlike
>README clearly states it is compatible and tested with Ubuntu 18.04 and specific CUDA version
>moron ignores this
>README explains how to set up the environment
>moron ignores this
>hurrr durrr it no work
Many such cases.
just use pip-tools with hashes.
why wouldnt you just take the modules out of the library directory and just put them in the project root at that point?
>we need VM because we are too moronic for a package manager
>third party paid "environment" manager
>docker just to run a scripting language
holy frick the python brainrot is real
>holy frick the python brainrot is real
hey I'm not defending python, I love how much the community has created thanks to it, look at all the neat AI apps but it definitely could be better. Anconda is paid so I guess if you're a business this is a no go but for the average g man going FOSS, it's fine to use and I recommend it. Love how it just bundles everything u need and conda just WORKS better than anything else. Docker really makes it painless running whatever the project is on different computers so the tradeoff of the hassle there I feel is okay. There aren't much better alternatives, otherwise those would had been more popular right now.
you're not "just running a scripting language"
>can't pip install in a virtualenv
HOLLLLLLLY SHIT HAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAAHAHAHAHHHAHHAHAA
Python dependencies are moronic just like the rest of the language for example GIL and how now these deluded Black folk start to realize a type system is in fact a good thing to have in your language.
usually changing the version in the requirements file works to be tbh fampai
Imagine being filtered by one of the most newbie friendly languages
My man, just chose another carer. You are not going to make it.
there is nothing easy about python except not having to manually manage memory
but ofc it's easy to show that you can print hello world in one line and without semicolons!
it attracts average cnile because that's all what they do (just without explicit memory management)
if you functional programming or oop in it you just cry
jesus fricking christ dude, if a numpy version mismatch sends you crying to fricking bot just drop everything and get a mcjob instead
>tried to screw together in a piece of ikea furniture
>pick up screw
>"size mismatch: screw is to big for the hole"
>instead of picking up the correct screw i go crying to an anonymous imageboard
"we are adults", says ikea, what a fricking joke!
Anon, Ikea systems has backwards compatible screws...
i guess me going to his repo made him see it and he commited a fix today? neat
Java and Matlab just works.
But please do use your toy language to have more headaches.
Are you too moronic to follow the instructions to use conda? It seems you are. Why don't you instead wait for them to make you an iOS app?