Skip to content
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

Cannot install ml-metadata v1.0.0 #161

Closed
AlexandreBrown opened this issue Jul 20, 2022 · 15 comments
Closed

Cannot install ml-metadata v1.0.0 #161

AlexandreBrown opened this issue Jul 20, 2022 · 15 comments

Comments

@AlexandreBrown
Copy link

Description

Hello, I cannot install v1.0.0 of ml-metadata using pip even tho the version seems available on the PyPi page : . https://pypi.org/project/ml-metadata/#history
Screenshot from 2022-07-20 07-17-46

Use Case

  • Use ml_metadata to acces Kubeflow artifacts information stored in the MySQL db.
@codesue
Copy link

codesue commented Jul 22, 2022

If you're using an Apple M1 chip, this could be related to #143.

@AlexandreBrown
Copy link
Author

AlexandreBrown commented Jul 22, 2022

@codesue Interesting, I have an AMD CPU (normal x86 architecture cpu) but I feel like for me the issue is finding the package on PyPi when doing a pip install.
Could you paste in your pip command ? Maybe there are flags that I'm not aware of to find the package.

@codesue
Copy link

codesue commented Jul 22, 2022

@AlexandreBrown I'm doing a regular pip install with the default resolver and no flags, e.g. pip install ml-metadata==1.9.0.

On my machine with an M1 Pro chip using Python 3.9.7, this produces the following error:

ERROR: Could not find a version that satisfies the requirement ml-metadata==1.9.0 (from versions: 0.12.0.dev0, 0.13.0.dev0, 0.13.1.dev0)
ERROR: No matching distribution found for ml-metadata==1.9.0

Installation works as expected on my machine with an Intel chip.

@AlexandreBrown
Copy link
Author

@codesue Ok, yes installing the latest version works for me as well being on a x86 architecture cpu, can you try to install v1.0.0 on both your machines to see if you have the same issue as me?

@codesue
Copy link

codesue commented Jul 22, 2022

@AlexandreBrown Ohhh, 1.0.0 doesn't install on either. Not related to #143.

@BrianSong
Copy link
Collaborator

@AlexandreBrown Thanks for bringing this up. What version of pip are you using?

@AlexandreBrown
Copy link
Author

AlexandreBrown commented Jul 29, 2022

Hello @BrianSong , I'm using pip version 21.2.4 (was also able to reproduce on latest version 22.2.1)

@BrianSong
Copy link
Collaborator

BrianSong commented Aug 5, 2022

Hi @venkat2469, I wonder does release engineer have any idea why 1.0.0 is not available through pip although it is available on the PyPi page: https://pypi.org/project/ml-metadata/#history?

@AlexandreBrown
Copy link
Author

@venkat2469 @BrianSong Hello, any update on this ?

@rtg0795
Copy link
Collaborator

rtg0795 commented Aug 16, 2022

Hi @AlexandreBrown , may I know the python version you are using to install the ml-metadata package?

@AlexandreBrown
Copy link
Author

AlexandreBrown commented Aug 16, 2022

Hello @rtg0795 , using Python 3.9 under Ubuntu (tried various Ubuntu versions, 18.04, 20.04, 22.04).

@BrianSong
Copy link
Collaborator

@AlexandreBrown , Python 3.9 is not supported until 1.8.0. Could you try to install 1.8.0 or 1.9.0?

@AlexandreBrown
Copy link
Author

@BrianSong Unfortunately 1.8 and 1.9 cause errors. I must install a version that is compatible with Kubeflow (v1.4.1).

Kubeflow 1.4.1 uses a database scheme that seems to only work with lower versions like 1.0.

My use case is I want to use ml-metadata to query the mysql database that Kubeflow fills in during experiments (pipeline runs).

Screenshot_20220816-161926.png

@BrianSong
Copy link
Collaborator

BrianSong commented Aug 16, 2022

I see. Is it possible for you to upgrade Kubeflow with a higher version? v1.4.1 is a pretty old version so it might not be compatible with MLMD 1.8.0 and higher.

Also, downgrading your py version looks like another way to go.

@AlexandreBrown
Copy link
Author

AlexandreBrown commented Aug 18, 2022

@BrianSong Well to be quite honest I wouldn't consider Kubeflow v1.4.1 as "old" since the latest version is v1.5.1 which was published only 13 days ago from AWS and upgrading Kubeflow installs can involve much more than a pip install (unfortunately).

I followed your alternative option and downgraded my python version from 3.9 to 3.8 and that did the trick for me.

Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants