Download ML thing.
make new venv.
pip install -r requirements.txt
.
pip can't find the right versions.
pip install --update pip
.
pip still can't find the right versions.
install conda.
conda breaks for some reason.
fix conda.
install with conda.
pytorch won't compile with CUDA support.
install 2,000,000GB of nvidia crap from conda.
pytorch still won't compile.
install older version of gcc with conda.
pytorch still won't compile.
reinstall the entire operating system with debian 11.
apt can't find shitlib-1.
install shitlib-2.
it's not compatible with shitlib-1.
compile it from source.
automake breaks.
install debian 10.
It actually works.
"Join our discord to get the model".
give up.
Programmer Humor
Welcome to Programmer Humor!
This is a place where you can post jokes, memes, humor, etc. related to programming!
For sharing awful code theres also Programming Horror.
Rules
- Keep content in english
- No advertisements
- Posts must be related to programming or programmer topics
It feels like you stood behind me yesterday, taking notes.
This comment gives me ptsd
thats when you do
/usr/bin/python3.11 -m pip install
mother. fucking. hardcoded paths. 1 step forward, 10 steps backward.
I recommend distrobox for adhoc distrohopping. Though for Nvidia stuff it links to drivers and cuda that you have installed on your host, so... I recently needed cuda 11.8 and that was hella fun to get going.
It's like my own little Fediverse
Feels very validating to see that everyone else's Python is held together by a thread too.
My workflow:
cd project
python -m venv .venv
. ./.venv/bin/activate
pip install -e .
By default pyvenv excludes system packages, so I can have different versions in the venv. To reset the venv, I just have to delete the .venv dir.
I've been using pipenv for a good while but I've started to move over to venv slowly, and I like it so far. It's a bit more of manual work but I feel like it's worth it.
i've moved to just using conda environments. i find it's a lighter load on my old brain.
I love this workflow because it has only two prerequisites: python and pip. It works on windows, linux, any vm or container. Pipenv requires some setup, while this should work everywhere. In powershell you have to use ./.venv/bin/acticate.ps1
but that's the only difference.
What did you not like on pipenv in comparison to venv? I was always avoiding venv because it was, as you said, manual work and it was too much effort to again google what was the order of commands and parameters to start a venv, which is not an issue in pipenv, since you just pipenv install what you need.
We had some issues in our CI where pipenv would sometimes fail to sync. It has recently gotten better, I think due to a fix of some race condition due to parallel installation. I think venv
would be better suited for CI in general, since it allows the use of a simple requirements.txt
file.
The other thing is I think it is rather slow, at least on windows which most of my team uses.
To conclude, I think as long as you aren't having any trouble and it simplifies your environment, you might as well use it.
Seconded, after being burned repeatedly I always do this. But why are you calling activate from the directory above?
I don't think he does. If you're talking about the third line, there's a space between the dots.
The dot command is equivalent to source
(running the script in the context of the current shell).
Exactly, venv solves this issues for me
Thank god for NixOS. (My daily on my laptop, seriously flakes + nix-direnv
is godsend for productivity. Reliable development environments and I don’t have to lift a finger!)
Do you have any troubles running it as your daily OS? Do you use it as your hobby or also for your work?
I know Nix and use it as my package manager, but I'm not sure about the experience with NixOS. So I'm still reluctant to make the switch.
I agree. I use a Mac but use nix to manage all this mess.
And the vscode direnv extension just makes it all work together.
I personally use Neovim (it's not nearly as much work as people make it out to be), so it's all integrated within my terminal.
I've recently discovered pipenv, and it has been a massive QoL improvement. No need to figure out bazillion of commands just to create or start an environment, or deal with what params should you use for it like you do with venv. You just pipenv install -r requirements.txt, and everything is handled for you. And when you need to run it, just pipenv run python script.py and you are good to go.
The best thing however are the .pipfiles, that can be distributed instead of requirements.txt, and I don't get why it's not more common. It's basically requirements, but directly for pipenv, so you don't need to install anything and just pipenv run from the same folder.
Yessssss
I actually wrote a script to make a folder an instant pipenv environment for me. Add it to your ./.zshrc. Has saved me a ton of time, I just removed some spaghetti lines that would reinstall pip and shit because it's when I was still early days into Py dev, now I work more with Py than I do C# and I'm a senior C# engineer, I just enjoy the masochism of py.
Also added a check for Arch/Ubu.
# Automated python virtual environment.
#######################################
VENV(){
if ! [ -x "$(command -v pipenv)" ]; then
echo "pipenv not installed... installing it now!"
sudo pip install pipenv
OS="$( ( lsb_release -ds || cat /etc/*release || uname -om ) 2>/dev/null | head -n1 )"
if [[ "$OS" == *"buntu"* ]]; then
sudo apt install pipenv -y
elif [[ "$OS" == *"rch"* ]]; then
sudo pacman -S pipenv
fi
pip install pipenv --upgrade
echo "Installation complete!"
fi
if [ -n "$1" ]; then
echo -e "Args detected, specifically using version $1 of python in this project!"
version="$1"
else
version=$(python -V)
version=$(echo "$version" | sed -e 's/Python //')
if [ -z "$version" ]; then
version=$(python3 -V)
if [ -z "$version" ]; then
echo "No python version installed... exiting."
return
fi
fi
fi
echo -e "\n===========\nCreate a Python $version virtual environment in $PWD/.venv [y/n]?\n==========="
read -r answer
case $answer in
[yY][eE][sS]|[yY])
export PIPENV_VENV_IN_PROJECT=1
pipenv --python "$version"
pipenv install -r ./requirements.txt
echo -e "\n\n\nVirtual python environment successfully created @ $PWD/.venv!\n"
echo -e "To run commands from this dir use 'pipenv run python ./main.py'"
echo -e "To enter a shell in this venv use 'pipenv shell'."
echo -e "To install from a requirements text file use 'pipenv install -r requirements.txt'"
echo -e "To update pip + all pip modules use 'pipenv update'!\n"
echo -e "Additional information can be found @ https://pipenv-fork.readthedocs.io/en/latest/basics.html"
;;
[nN][oO]|[nN])
echo "Fine then weirdo why did you run the command then, jeez.Exiting"
;;
*)
echo "Invalid input..."
;;
esac
}
I could redraw this whole chart using only references to pipenv based on my experiences with managing it alongside other tools (especially homebrew). It’s good at many things but is no magic bullet.
Yeah but is it really worse than python3-venv like some people act like it is? I just don't see it.
Now take all of this and find something that needs an old out of date Python... Like 2.6.
.. cry later
Oh, so that is where all of my modules are? That makes total sense, thanks!
Throw pyenv in there and add some more complexity!
Hopefully Mojo will sort it all out. Maybe even inspiring a new, positive streak of xkcd strips in the future?
This is now the process to configure your environment for learning Python through Minecraft Java.
Honestly, at this point I'm running all my python environments in different docker containers. Much easier to maintain.
Wait, you guys don't have a vm for each project and just use ssh to work on them?
How does the workflow works in practice? You just use the containers to compile your code, or do you actually have a whole dev environment with IDE and everything and work directly in the container? I can't imagine how does the workflow looks. Or is it possible to set up i.e. a JetBrains Rider to always spin up a container to compile the code in it? But then, if all the requirements and libraries are only on the container, how would it be able to do syntax highlithing and Intelisense (or what's the correct work for code completion), if it doesn't have the libraries on the host?
I'm probably missing something, but all the solutions I can figure out with my limited experience have issues - working on IDE in a VM sounds like a nightmare with moving files between VM and host, and the whole "spin up a VM, which takes time and it usually runs slower on the shitty company laptop, just to make a quick edit in one project". And I feel like setting up an IDE to use environment that's in a VM, but the IDE runs on a host sounds like a lot of work with linking and mounting folders. But maybe the IDEs do support it and it's actually easy and automated? If that's the case, then I'll definitely check it out!
Check out dev container in VSCode. Even better with Codespaces from Github. You can define the entire environment in code, including extensions, settings, and startup scripts along with a Docker container. Then it's just one button click and 5 min wait until it's built and running. Once you have built it you can start it up and suspend it in seconds, toss it out when you don't need it, or spin up multiple at once and work on multiple branches simultaneously.
Part of that was a joke, but i do sometimes use a windows vm when i am working with a windows build, só i just have the vm openned on my second monitor and emacs open on my main machine, than i ssh within emacs into the vm and build and debug the code there, this way i don't need a whole dev enviroment, just git and the build tools
I've been trying to sell this idea to my team for a year now. I've even done all the legwork in my free time with a personal project and I've offered the patterns to the team. But alas, we still commit to masochism.
Agreed! That's the way we do work projects. For personal stuff, I also like using pyenv. But yeah, that's it, keep it nice and simple.
phew, glad to hear I'm not a total idiot. 👍
Poetry is nice for this but honestly, Rust's cargo and the JS npm/yarn have spoilt me.
As a mac user I feel this
This Helen’s on my laptop from following AI install scripts. I high key hate Python, it’s my hated language and I wish another language was the default for ML.