You are not logged in.
Hi all,
There is a bug in tensorflow and the suggested workaround is to use h5py < 3.0.
The community/python-h5py version is 3.1.0 and `pip install h5py<3.0` seems to be incompatible with community/hdf5 1.12.0-2 because it receives a compilation error.
Any ideas what I can do?
Offline
Downgrade until fix applied?
How bad is your situation making that approach unfeasible or unwanted?
Offline
Downgrade python-h5py? Unfortunately that reverts it to a python 3.8 installation and downgrading Python does not seem feasible.
I actually found a way after posting this:
1) Downloaded h5py 2.10.0 from source
2) Downloaded the 2 patches that were removed from the arch git for python-h5py https://github.com/archlinux/svntogit-c … 7951/trunk
3) Manually patched the h5py 2.10.0 source with these 2 files
4) Manually installed via `python ./setup.py install --user`
I think this will work for now but I'm still open to any... cleaner suggestions
Last edited by mcardillo55 (2020-12-05 03:00:01)
Offline
I have exactly the same problem. I cant downgrade h5py after Python was updated to 3.9.
I have no way now of loading trained models, since this gives me a h5py error.
Offline
@mcardillo55 Could you explain how you did it? I have the same problem
Offline