-
Notifications
You must be signed in to change notification settings - Fork 549
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[BUG (RISK?)]: v2 relies on unsupported usage of scikit-learn #752
Comments
the last option isn't too bad IMO. As-is, we "define" (air-quotes) the following scalers:
AFAICT, the only reason we use the n, p = 10, 4
X = np.random.rand(n, p)
scaler = StandardScaler()
scaler.mean_ = X.mean(0)
scaler.scale_ = X.std(0) so all that would mean in practice is saving the train statistics and just setting the attributes of a new |
For now, I think it is fine to bump this to v2.1. It currently raises a warning, so at least users can decide for themselves in the mean time. Additionally, we see this in the CI because developers have different version locally than the github actions use with a fresh install. I think it would be good to create the checkpoints with a fresh install once v2.0 is ready and the checkpoint files won't change anymore. |
@KnathanM @kevingreenman maybe we pin this issue so it's easier to find? |
Bug Risk
We have had a warning quietly being raised in our CI for a while now, like here, that says:
The way we load and unload the
StandardScaler
object is explicitly 'unsupported and inadvisable' in their docs.Path Forward
It's not immediately obvious what to do, though some options we have discussed online include:
Please chime in with your thoughts on this, v2 developers and users.
The text was updated successfully, but these errors were encountered: