-
Notifications
You must be signed in to change notification settings - Fork 255
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
Fresh install missing GMP #832
Comments
I found this issue which might be related: |
So this works out of the box on Ubuntu 20.04 LTS (If I just run |
I am using One other new thing I see is when I restart the kernel in jupyter, instead of killing the process, it responds with this: This is new as well. Thanks for your response; I will work on a clean room replication for this today and share it here. One simple issue I found is that for some easier problems, the above message is not generated. |
I am using automatic installation, just |
I am also getting these errors:
|
I replicated the error in a minimal setting:
env:
resulting in:
I can't share the actual problem I work on, but the gist of it is:
The problem has 1200 continuous 900 binary variables and 950 constraints. |
This is the complete log with
|
@xLaszlo There is an error, thanks for diving more into it! I never actually optimised the built model when I checked earlier, as I thought the error should be raised when I set the parameter. It is clear that GMP is just not linked properly. The intention was that GMP had been statically compiled and was also shipped via For some of your comments above:
@mmghannam It is going to be a good time to test the pipeline |
No worries. Thanks for taking a look at this. I must admit this is beyond my knowledge, so I don't think I can be much help. The only reason I mentioned the ctrl-c is that I didn't remember it happening before. I must admit this is an experimental project, and I am changing the algorithm and switching between solvers, so "Didn't happen before" might mean I just haven't yet used SCIP in that configuration. I asked ChatGPT how to check if GMP is on the machine, and all checks came back positive. If you have an update, I'd appreciate a comment here to be notified. Thanks again. |
Update: Apparently Boost is a hidden requirement for the extra precision. We're working on packaging it into Linux + Mac builds (Linux is currently working). Will be another week or so though as the build pipelines are failing for Windows. |
Describe the bug
I am getting the following warning:
Despite setting feastol to 1e-9. If I change the value to 1e-7, the warning disappears. Somewhere, it must be divided by 1000.
This started when I tried to reinstall the virtual environment, so I suspect it must have been the result of a recent change.
libgmp-dev is installed on the machine
To Reproduce
I am using cvxpy with a number of variables in the low thousands
This is the exact line:
problem.solve(solver=self.solver, verbose=self.verbose, scip_params={'numerics/feastol': 1e-9})
As can be seen, I am using 1e-9 for feastol.
Expected behavior
This was working before.
System
pyscipopt
?I tried both poetry and pip and installing it with
pip install pyscipopt
and alsopip install cvxpy[SCIP]
with no different results.The text was updated successfully, but these errors were encountered: