Skip to content
This repository has been archived by the owner on Apr 24, 2023. It is now read-only.

h5py encoding problem with pretrained models #18

Open
bertsky opened this issue Nov 25, 2021 · 1 comment
Open

h5py encoding problem with pretrained models #18

bertsky opened this issue Nov 25, 2021 · 1 comment

Comments

@bertsky
Copy link
Contributor

bertsky commented Nov 25, 2021

In a fresh installation with the current software ecosystem, I end up with the well-known issue

AttributeError: 'str' object has no attribute 'decode'

when loading the models. Essentially, older Keras/Tensorflow models cannot be loaded by h5py>2.10 (but are nevertheless dependent on that). Add pip's stupidity, and you can at least downgrade h5py manually, if that requirement comes last in the chain. But there's an additional glitch because h5py 2.1 does not compile with more recent Numpy anymore.

EDIT: confused 2.10 with 2.1, sry

@Zathiron
Copy link

Zathiron commented Jan 7, 2022

where you able to solve the issue? I am having the same problem and can't solve even if I have the correct versions.
Nevermind was solved using conda install 'h5py==2.10.0' in the anaconda env.

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

No branches or pull requests

2 participants