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

tensorflow 2.4 depends on h5py 2.10 but is packaged with h5py 3.2 #124599

Closed
sekunho opened this issue May 27, 2021 · 6 comments
Closed

tensorflow 2.4 depends on h5py 2.10 but is packaged with h5py 3.2 #124599

sekunho opened this issue May 27, 2021 · 6 comments

Comments

@sekunho
Copy link
Contributor

sekunho commented May 27, 2021

Describe the bug
h5py 3 is incompatible with tensorflow 2.4 (see here).

To Reproduce
Steps to reproduce the behavior:

  1. Load model
  2. Run script
  3. Encounter error
# Error when loading model
model_config = json.loads(model_config.decode('utf-8'))
AttributeError: 'str' object has no attribute 'decode'

Expected behavior
It's supposed to be packaged with h5py 2.10 to avoid the bug.

Notify maintainers

@jyp @abbradar

@nixos-discourse
Copy link

This issue has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/overriding-a-dependency-of-a-package/13307/1

@danieldk
Copy link
Contributor

danieldk commented May 27, 2021

The general issue is that we cannot have multiple versions of Python packages in nixpkgs (since multiple versions could then end up in the interpreter path). I think this is best resolved by updating Tensorflow in nixpkgs to 2.5.0.

@CMCDragonkai
Copy link
Member

Can we either:

  1. Upgrade TF to 2.5, and backport that to 21.05 release
  2. Downgrade h5py to version 2, and backport that to 21.05 release

At the moment TF on 21.05 release is broken due not keras model loading.

@sekunho
Copy link
Contributor Author

sekunho commented Jun 2, 2021

The restriction of being unable to have multiple versions for Python packages when nix supports it is a bit inconvenient, and unreliable. Since problems like this can appear in the future where a dependency has to be pinned to a specific version, but isn't. I don't really know how things are under the hood though. I'm fairly new to using nix for packaging things.

@CMCDragonkai
Copy link
Member

Upgrading to 2.5 TF is probably the best way. Hoping to look into this later.

@CMCDragonkai
Copy link
Member

The 2.5 MR was merged by accident, this is still not resolved.

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

No branches or pull requests

5 participants