-
Notifications
You must be signed in to change notification settings - Fork 5
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
Library importation bug in GCloud VM #29
Comments
Hi @ProfLeao ... thank you for reporting. As a workaround, you can install cantera from the |
Thanks @ischoegl Worked perfectly. |
I was able to replicate this problem on Ubuntu 20.04. It seems the problem is that the Cantera extension module is linked specifically to
I looked and saw that all the Python packages installed in my conda environment explicitly specify |
Fwiw, I ran into the exact same issues for
on Fedora. The dev version (still 2.6.0b2) is obsolete, so I didn’t investigate further. |
It's not clear to me how the tests can pass on CI and this doesn't work on user machines. I guess Conda is solving for the 2022 version of mkl on CI but 2021 ends up in the user environment... Do you have a sense for what requires |
It's required by numpy. Here is the info for the version of numpy that currently gets installed into a clean environment with Cantera:
Note the explicit version limits on MKL. The version that's being used for at least some of the conda builds is:
Which doesn't block MKL 2022.0.1, and that's what gets installed. |
Is there a way for us to just depend on a shim library that links to either OpenBLAS or MKL, and doesn't require a specific version of either? The only functions we use are just standard LAPACK routines that are available in any implementation. I don't think it should be necessary for us have different packages for different versions of these libraries. |
|
@bryanwweber … good to know. I’ll pass on this one as I got stuck with MKL before (Cantera/cantera#1316) … fwiw, #33 temporarily disables MKL, although it should definitely be enabled again. |
MKL is disabled as a temporary fix of Cantera#29 to resolve Cantera#31. It should be reactivated in a permanent fix.
* Prevent installation of cantera from PyPI * Disable MKL MKL is disabled as a temporary fix of #29 to resolve #31. It should be reactivated in a permanent fix. * Add yaml2ck script * Update pytest Starting with Cantera 3.0, the Python test suite is no longer packaged with the Python module. This commit ensures that conda is still able to run the tests. * Replace deprecated SCons option * Include share/cantera/data in libcantera
Recent builds of Numpy seem to have separate builds for different MKL versions. Partial output of
So, one build that specifically requires MKL 2021 and one that specifically requires MKL 2023, in addition to a build for OpenBLAS (and a conspicuous absence of a build for MKL 2022, which was the source of some of the original trouble reported here). I think for us, all the extra builds really aren't worth it, and we should just stick to using OpenBLAS for the packages on the |
Superseded by #46 |
I've been experiencing the following error when importing the library when using it in a gCloud VM:
The enviroment has following setup:
The OS is 20.04.1-Ubuntu.
I appreciate any help.
The text was updated successfully, but these errors were encountered: