Re: Problems with Linear Elasticity Example
On 05/09/2017 01:36 AM, Jeshwanth.Kundem@wagner-group.com wrote:
You are right about scikit-umpfack being installed somewhere else. I removed it and it worked as you mentioned in the previous email.
OK. Now if you do a fresh scikit-umfpack install, does it help?
I will check the example.
I have to disagree with the Anaconda statement. I tried, tried and tried with Anaconda. It might be an easy install, but what is happening is the Anaconda is at 4.3.1. When I install sfepy through conda forge channel, the sfepy or conda forge wants the anaconda to be 4.2.13 or something. For that reason, it gave me an option mentioning "The packages will be superseded by a higher priority channel" and it is actually trying to get back to 4.2.13. Because of this lot of anaconda packages are being unlinked with conda. They are causing other problems down the line, with mayavi, wxpython, serial etc. I would like to know what you think about my statement.
I am not aware of any such anaconda version dependence in the sfepy recipe - maybe some of the dependencies need that. Anyway, I am not sure how to solve that. If you happen to find some more information, let us know.
Are you mentioning that if I use conda, I would not have the scikit-umfpack problem and UMFPACK would work without any errors? If so let me know, I will give it a try once again.
It might. I did not try it recently, but that was the idea. If it does not work, submit an issue to the package that fails, please.
r.
Because of this I intentionally deleted anaconda and installed python and packages one by one with pip.
-----Original Message----- From: Robert Cimrman [mailto:cimrman3@ntc.zcu.cz] Sent: Monday, May 08, 2017 5:03 PM To: Kundem, Jeshwanth Cc: sfepy@python.org Subject: Re: [sfepy] Problems with Linear Elasticity Example
On 8.5.2017 23:34, Jeshwanth.Kundem@wagner-group.com wrote:
Hi Robert
The example works when I change the ls solver from scipy_direct to scipy_iterative. Not when I uninstall scikit-umfpack. It is apparently not being defaulted to scipy LU solver. Any thoughts on this? However, I am
If you remove scikit-umfpack and still see the same error, I would guess you have another version of scikit-umfpack installed somewhere.
unable to specifiy the number of iterations if I change to scipy_iterative.
See examples/diffusion/cube.py for an example of how to set the number of iterations.
The UMFPACK has to be installed before using sckit-umfpack. There is no good documentation on that . I got onto this website: http://faculty.cse.tamu.edu/davis/suitesparse.html . However it is only available for MATLAB and C.
How did you install python and other sfepy dependencies? (Which versions?) The easiest way nowadays is to use anaconda - there is even a recipe for sfepy on conda-forge ("conda install -c conda-forge sfepy" should work).
r.
participants (1)
-
Robert Cimrman