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
gmpy prerequisite is broken on windows systems #452
Comments
I'm looking to an alternative to gmpy at the moment. Will keep you posted. |
FYI, gmpy2 2.0.8 has available pre-built wheels for recent python versions (not on pypy, and I don't understand why they don't mirror them over there). I don't know if that could be enough, but I suppose that migrating from gmpy to gmpy2 should be fairly straightforward (I haven't seen anything about that, nor I know\understand the purpose of this module, but I hope they have a similar shape\size :)) |
Unfortunately it is not straightforward as the functions I'm using are not available anymore in gmpy2. But don't worry the solution that I have working is a simple pure python alternative, will land in a couple days at most. |
Thank you for your updates! Unfortunately, it still appears that gmpy is in the requirements list or somewhere in the ludwig install. On windows 10, pip install ludwig outputs all the requirements already satisfied and this: |
Install from master. |
On windows systems,
pip install gmpy
is broken, alas it needs a lot of prerequisites to be built correctly (first error message you get, is that the wheel cannot be built due to missing gmp.h, for anyone else out there trying to search for this issue, and on windows we don't have libgmp3-dev to install).Now a "simple"
pip install ludwig
now fails due to missing header files and more.I know you are not explicitly supporting Windows installs, but up to the H3 feature support (when gmpy was added) the package was able to be built\deployed.
I suppose it could be possible to
pip install ludwig
up to the point where all other packages are correctly installed and then do a finalpip install ludwig --no-deps
, and then manually commenting all H3 references inludwig\features\feature_registries.py
but it's a really broken way to workaround over this.(in my case, with an added
pip uninstall tensorflow
andpip install tensorflow-gpu
to leverage GPU processing)Preliminary tests show that this "half-and-half" workaround with --no-deps is running fine with train\test\predict\visualize operations (of course not including anything relative to H3). I am also seeing that 0.1.2 models cannot be used with 0.2 due to
Not found: Key optimizer/intent/fc_0/biases/Adam not found in checkpoint [[{{node save/RestoreV2}}]]
I am looking into this to find if it's related to H3 commenting or if it deserves a separate bug report (even just to give a readable message stating the non retrocompatibility of 0.2 towards 0.1.2)
Thanks!
The text was updated successfully, but these errors were encountered: