ci: speed up by using uv to install python packages#2082
ci: speed up by using uv to install python packages#2082afuetterer wants to merge 1 commit intoMaartenGr:masterfrom
Conversation
|
I love the rustification of all these packages! Haven't used Having said that, when we test using |
|
If you look at a car or a bike, rust is usually a bad thing. But in this case, it is really great with the massive performance improvements. Actually I am not sure about your question. Your package requires I don't know, its your call. If users use |
It might be a problem if one of those versions results in bugs that we haven't tested through the unit tests. If
That's true, the speed-up is definitely there. I am just wondering how many users will actually be doing this. In my experience, >90% of users will just run I'm gonna be honest with you here, I'm not sure what the chance actually is that the testing pipeline using |
|
Yes, sure, things can break, if the API of a dependency changes, and there is no upper bound in place. It's your decision. Happy to close this, if not desired. Just wanted to show the difference in installation time. |
|
Let's close it for now and re-open when we run into issues with the speed up of the pipeline or if |
|
Sure, no problem. |
|
I saw that you now have uv installation instructions at the top of the readme. Would you like to give this one another try? uv has matured quiet a lot in the last year and has lots of users. There is also on official uv GHA, that could be used in CI. https://github.com/astral-sh/setup-uv Please let me know what you think. I can create a new PR, if thats okay. |
|
Definitely agree! You were right to suggest it; BERTopic needed some changes first to actually be able to use |
|
Should I open a ticket to discuss further or just go for a new PR then? |
|
@afuetterer No new ticket is needed; we already discussed it here. |
|
Just making sure. |
What does this PR do?
This is a proposal to use the uv for installing python packages in CI. It concurrently downloads the packages from pypi and improves the time of the install step significantly. From 3+ minutes to 30+ seconds.
What do you think?
Before submitting