Minor edits to benchmarking code by quaquel · Pull Request #1985 · mesa/mesa

quaquel · 2024-01-21T10:22:37Z

fix for < python 3.12
Insert the current repo of mesa into sys.path to ensure you are using the correct version
pep8 name changes of files
added code to run each benchmark in isolation (useful for profiling and line profiling, used it extensively already for Proposal: Formal neighborhood definition for grids #1900 (comment))

for more information, see https://pre-commit.ci

EwoutH · 2024-01-21T13:45:25Z

Let's see if the CI is working: /rerun-benchmarks

.gitignore

github-actions · 2024-01-21T13:50:35Z

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
SchellingModel	small	🔵 -2.8% [-3.2%, -2.4%]	🔵 -0.1% [-1.1%, +0.7%]
SchellingModel	large	🔵 -2.7% [-3.2%, -2.2%]	🔴 +5.4% [+3.8%, +7.0%]
WolfSheep	small	🔵 -0.4% [-1.0%, +0.1%]	🔵 +0.6% [+0.3%, +0.9%]
WolfSheep	large	🟢 -27.9% [-42.8%, -13.4%]	🔵 +1.2% [-1.9%, +4.4%]
BoidFlockers	small	🔵 -0.9% [-1.2%, -0.5%]	🔵 -0.5% [-1.3%, +0.2%]
BoidFlockers	large	🔵 -0.1% [-0.7%, +0.6%]	🔵 -0.5% [-1.1%, +0.1%]

EwoutH · 2024-01-21T14:18:30Z

Let me know when this is ready for review

commit c0de4a1 Author: Ewout ter Hoeven <[email protected]> Date: Sun Jan 21 14:43:08 2024 +0100 Add CI workflow for performance benchmarks (mesa#1983) This commit introduces a new GitHub Actions workflow for performance benchmarking. The workflow is triggered on `pull_request_target` events and can be triggered with a "/rerun-benchmarks" comment in the PR. It should be compatible with PRs from both forks and the main repository. It includes steps for setting up the environment, checking out the PR and main branches, installing dependencies, and running benchmarks on both branches. The results are then compared, encoded, and posted as a comment on the PR.

This reverts commit 9485087.

quaquel · 2024-01-21T18:00:37Z

Let me know when this is ready for review

It should be alright now. Please squash and merge this because I made a bit of a mess of the commit history.

EwoutH · 2024-01-21T18:03:00Z

Could you update the PR message with a bit of motivation for the changes made?

/rerun-benchmarks

quaquel · 2024-01-21T18:07:45Z

Could you update the PR message with a bit of motivation for the changes made?

/rerun-benchmarks

I already did so, or do you want to know more? But if so, what?

github-actions · 2024-01-21T18:07:46Z

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
SchellingModel	small	🔵 -2.4% [-2.8%, -2.0%]	🔵 -0.3% [-0.9%, +0.2%]
SchellingModel	large	🔵 -1.9% [-2.5%, -1.3%]	🔵 -0.0% [-0.7%, +0.7%]
WolfSheep	small	🔵 -0.0% [-0.3%, +0.2%]	🔵 +0.0% [-0.1%, +0.1%]
WolfSheep	large	🔵 -4.6% [-13.7%, +0.6%]	🔵 +0.7% [+0.2%, +1.3%]
BoidFlockers	small	🔴 +3.6% [+3.3%, +4.0%]	🔵 +0.4% [-0.2%, +0.9%]
BoidFlockers	large	🔵 +2.9% [+2.5%, +3.3%]	🔵 +0.3% [-0.2%, +0.8%]

benchmarks/WolfSheep/__init__.py

github-actions · 2024-01-21T18:13:07Z

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
SchellingModel	small	🔵 -1.0% [-1.5%, -0.5%]	🟢 -5.7% [-6.4%, -4.9%]
SchellingModel	large	🔵 -1.2% [-2.0%, -0.2%]	🟢 -5.8% [-6.4%, -5.1%]
WolfSheep	small	🔵 +1.3% [+0.9%, +1.6%]	🔵 -1.2% [-1.3%, -1.0%]
WolfSheep	large	🟢 -21.0% [-37.0%, -5.3%]	🔵 +2.9% [-3.1%, +9.2%]
BoidFlockers	small	🔴 +3.8% [+3.2%, +4.4%]	🔵 +1.2% [+0.6%, +1.8%]
BoidFlockers	large	🔵 +3.1% [+2.0%, +3.8%]	🔵 +0.8% [+0.0%, +1.6%]

EwoutH · 2024-01-21T21:16:26Z

LGTM, thanks for this effort. @rht or @Corvince could one of you do the final review and squash merge if ok?

Did you also observe the Schelling being faster on your system? Or is that an anomaly?

In the long run we should try to keep these models in sync with ABMFrameworksComparison. Do we want to make further improvements on these models in the short run?

rht · 2024-01-21T23:41:04Z

In the long run we should try to keep these models in sync with ABMFrameworksComparison. Do we want to make further improvements on these models in the short run?

The sync with Agents.jl devs needs to happen soon, if possible. I think they would appreciate to know if we are now monitoring the performance as well.

rht · 2024-01-21T23:43:48Z

Did you also observe the Schelling being faster on your system? Or is that an anomaly?

Haven't tested, but the finding is probably not a blocker to this PR being merged.

quaquel · 2024-01-22T06:47:31Z

LGTM, thanks for this effort. @rht or @Corvince could one of you do the final review and squash merge if ok?

Did you also observe the Schelling being faster on your system? Or is that an anomaly?

In the long run we should try to keep these models in sync with ABMFrameworksComparison. Do we want to make further improvements on these models in the short run?

I think there is still quite some noise in the benchmarks. But one issue with the old schelling code was that it did not use self.random properly in all places. This probably contributes to even more noise.
In my view we should indeed keep our versions in sync with ABMFrameworksComparison. It would be good to push any changes to them in the context of a new stable release of MESA.

I also noticed that, in particular, WolfSheep is not a good Python implementation. There is a lot of unnecessary code duplication going on. I have the impression that some implementations were done to mimic NetLogo's version. Which is fine for a tutorial for people coming from NetLogo. However, I would argue that the benchmark (and most examples) should reflect the Pythonic and Mesa way of doing things. I am happy to push a WolfSheep reimplementation if desired (But also see https://github.com/quaquel/mesa/blob/cell-space/benchmarks/WolfSheep/wolf_sheep.py, for the direction in which I would change it.

EwoutH · 2024-01-22T07:04:04Z

Sounds good, would you like to do both (pushing to ABMFrameworksComparison and speeding up WolfSheep yourself?

quaquel · 2024-01-22T07:29:49Z

Yes, I'm fine with doing both (although it is more about cleaning up wolfsheep code; I have no idea about performance, but it is probably small).

rht · 2024-01-24T05:25:57Z

I think the first step is to simplify the WolfSheep at https://github.com/projectmesa/mesa-examples, then next all the other WolfSheep's (in Mesa's benchmark and Agents.jl's benchmark) should be updated to be almost the same as this reference implementation.

quaquel and others added 9 commits January 15, 2024 11:23

Make RandomActivationByType.agents_by_type backward compatible

bc1718c

[pre-commit.ci] auto fixes from pre-commit.com hooks

a7ddad5

for more information, see https://pre-commit.ci

updated warning

1f9af49

Merge remote-tracking branch 'upstream/main'

59885b6

Merge remote-tracking branch 'upstream/main'

b9fe806

Update .gitignore

2076b6d

Merge remote-tracking branch 'upstream/main'

600c372

Update global_benchmark.py

c83179d

pep8 rename

4c8cb8b

Update schelling.py

7bccdd1

EwoutH reviewed Jan 21, 2024

View reviewed changes

.gitignore Outdated Show resolved Hide resolved

Update schelling.py

64d8c46

quaquel added 2 commits January 21, 2024 14:57

Update schelling.py

1e794d3

Update schelling.py

4bc5fac

quaquel added 4 commits January 21, 2024 17:53

Revert "Squashed commit of the following:"

5cd0e74

This reverts commit 9485087.

Update .gitignore

63eec5e

further updates

2a4459c

quaquel commented Jan 21, 2024

View reviewed changes

benchmarks/WolfSheep/__init__.py Outdated Show resolved Hide resolved

Update benchmarks/WolfSheep/__init__.py

1b5661e

EwoutH approved these changes Jan 21, 2024

View reviewed changes

EwoutH requested a review from rht January 21, 2024 21:11

EwoutH requested a review from Corvince January 21, 2024 21:11

rht merged commit 4d0eeec into mesa:main Jan 21, 2024

Uh oh!

Conversation

quaquel commented Jan 21, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

EwoutH commented Jan 21, 2024

Uh oh!

Uh oh!

github-actions bot commented Jan 21, 2024

Uh oh!

EwoutH commented Jan 21, 2024

Uh oh!

quaquel commented Jan 21, 2024

Uh oh!

EwoutH commented Jan 21, 2024

Uh oh!

quaquel commented Jan 21, 2024

Uh oh!

github-actions bot commented Jan 21, 2024

Uh oh!

Uh oh!

github-actions bot commented Jan 21, 2024

Uh oh!

EwoutH commented Jan 21, 2024

Uh oh!

rht commented Jan 21, 2024

Uh oh!

rht commented Jan 21, 2024

Uh oh!

quaquel commented Jan 22, 2024

Uh oh!

EwoutH commented Jan 22, 2024

Uh oh!

quaquel commented Jan 22, 2024

Uh oh!

rht commented Jan 24, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

quaquel commented Jan 21, 2024 •

edited

Loading