-
Notifications
You must be signed in to change notification settings - Fork 0
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
Optimization #19
Comments
I add documentation instructions to |
I'm going to store various loose ends to come back to that pertain to Z3 optimization is this issue, the first of which was from when I was experimenting with different non-trivial proposition constraints.
Here are some examples where the
|
Coming here after raising the no quantifiers issue for optimization—I think removing |
It'll be easy to remove |
A potential workaround is just to remove existential quantifiers except when universal scope is needed over an existential; the universal > existential constraints are only for the hyperintensional operators that aren't |
Hey that's a good shout! Graham also suggested eliminating quantifiers (defining them in python) since the models are all finite and small. Once this has been done, I was thinking of adding a flag which overrides One reason to avoid removing I'm close to finishing a new draft on the paper for this project (same content though the narrative has changed a lot), but will return to some optimization (raising issues on the Z3 GitHub) once I do. Hope that your summer is going well! |
Sounds good! I'm working on the python definitions now, so far looking alright—harder than I thought but I may just be overthinking it. Looking forward to seeing the new draft, and hope your summer's going well too :) |
Made a new version of I tested it with premises = ['(A \\boxright C)', '(B \\boxright C)']
conclusions = ['((A \\wedge B) \\boxright C)'] Turning on both
Hopefully the model is right and |
I ran every test in The tests that should not find models that work with |
That's great! I'll take a look at the code later today. Looking at the example you included above, note that the premises are both false. I ran into this issue while adding other operators to the language. I am not totally sure what makes this happen but guessed that it is where the constraints being generated from the input sentences aren't quite right. That is, the semantics generate something, but not the right things for the premises to be true and the conclusion false (I labeled these false premise models in the notes). I have also run into cases where the premises are true and the conclusion is also true (calling these true conclusion models). By improving the semantics for the new operators, I was mostly able to eliminate these issues. This might be another case where looking at all generated constraints may be helpful to see what is going on. I'll try to take a look at this soon. As for the imports, I am yet to find someone who knows how to set this up correctly, and have found lots of conflicting information online, no combination of which has worked so far. But I hope to straighten that out soon and will document it accordingly. |
I tried running tests on the new branch but couldn't get it to work. Setting
I was wondering if |
Huh that's interesting, can you show what the entire error output is? |
|
Ran into the same error running |
I've been trying to fix the problem but can't seem to sort it out. Seems to be an issue with
It seems this has the same issue, where the problem seems to be with |
I've added benchmarking to
test_complete.py
in order to begin comparing results. Here is how the proposition constraints were originally defined:This ran with an execution time of .71 sec. I then replaced this constraint with:
This ran with an execution time of .18 seconds. This makes me curious what kind of improvements can be gained by removing all occurrences of
Exists
from the Z3 constraints generated by the functions insemantics.py
, replacing these with unique variables instead.The text was updated successfully, but these errors were encountered: