https://flyte.org logo
#ecosystem-unionml
Title
# ecosystem-unionml
a

Alain GALDEMAS

11/27/2023, 3:13 PM
Hello, I face an issue, with the basic sample app.py from unionml, when I try to follow the documentation : https://unionml.readthedocs.io/en/stable/flyte_cluster.html#unionml-predict my flyte target is a flyte-binary deployed on an ovhCloud kubernetes cluster. I can deploy and run some simple workflows with flytectl & pyflyte. I can execute
unionml train app:model -i '{"hyperparameters": {"C": 1.0, "max_iter": 1000}}'
, and the workflow execute and run OK in flyte, but when I want to predict
unionml predict app:model -f data/sample_features.json
I've got an error :
TypeError: No automatic conversion found from type <class 'sklearn.linear_model._logistic.LogisticRegression'> to FlyteFile.Supported (os.PathLike, str, Flytefile)
which is a flytekit originated error, my config is unionml 0.2.1, flyteidl 1.5.21, flytekit 1.9.1, scikit-learn 1.3.2, with flyte-binary I use the app.py created with unionml init project, with a remote configuration towards the target flyte-binary, which seems ok I'm stuck with this issue, any idea on how to fix, by specifying a type somewhere ? is there a change to do in the app.py, to avoid this error ? is it a bug due to some new version of packages ? thanks in advance for any help to fix this iissue
n

Niels Bantilan

11/30/2023, 6:11 PM
hi @Alain GALDEMAS gonna try to repro later today, but in the mean-time, can you try it with
flytekit==1.8.1
? Flytekit itself has undergone many changes and wanted to see if it supports the minimum pinned version
a

Alain GALDEMAS

12/01/2023, 8:17 AM
ok, I'll try asap and come back to you !
I downgraded flytekit to 1.81 (
sudo pip install -Iv 'flytekit==1.8.1
), unfortunately I got the same behavior and error:
unionml predict app:model -f data/sample_features.json
give:
TypeError: No automatic conversion found from type <class 'sklearn.linear_model._logistic.LogisticRegression'> to FlyteFile.Supported (os.PathLike, str, Flytefile)
should it be an issue with my own python environment under macos, related to the fact I have to sudo pip install, else I've got some file permission issues issue... ? I'll ask some of my colleagues to test from their machines ! on my side I'll try to test with a local flyte-sandbox, that should confirm an issue with my own local python environment... If you have some clue about my supposition, don't hesitate...