Skip to content
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

Benchmarking and improving HMC/NUTS #19

Open
5 tasks done
yebai opened this issue Jul 26, 2017 · 7 comments
Open
5 tasks done

Benchmarking and improving HMC/NUTS #19

yebai opened this issue Jul 26, 2017 · 7 comments

Comments

@yebai
Copy link
Member

yebai commented Jul 26, 2017

I suggest that we use this issue as an umbrella page for tracking improvements of benchmarking performances (memory, running time). For every new commit, we could write a small julia script to automatically (see GitHub.jl) post a new comment to this issue containing updated runtime information.

Related issues:

@yebai
Copy link
Member Author

yebai commented Jul 26, 2017

Model Turing Stan Ratio
Mixture-of-Categorical 21.31 5.93 3.59
Bernoulli Model 3.52 0.02 175.88
Binormal (sampling from the prior) 3.31 0.04 82.67
School 8 2.27 0.1 22.69
LDA 84.27 3.12 27.01
Simple Gaussian Model 3.1 0.02 154.89
LDA-vec 18.79 3.12 6.02
Kid 14.62 1.46 10.01

PS. perhaps we should link back to the commit as well.

@yebai yebai changed the title Benchmarking Benchmarking and improving HMC/NUTS Jul 26, 2017
@xukai92
Copy link
Member

xukai92 commented Apr 4, 2018

Link to Benchmark spread sheet

@yebai yebai transferred this issue from TuringLang/Turing.jl Mar 21, 2019
@xukai92
Copy link
Member

xukai92 commented Mar 25, 2019

I wonder if there is a GitHub service like Travis does this.

@yebai
Copy link
Member Author

yebai commented Mar 25, 2019

It's possible to perform these benchmarking using Travis, by running each task in a separate stage. For example

  • stage: LDA, run PR and master, compute runtime ratio
  • stage: Mixture: run PR and master, compute runtime ratio
  • ...

All these runtime ratios can be automatically posted into PR comments (Mohamed already implemented this in TuringBenchmarking), and saved in a separate repo.

Fully supporting this idea still requires some work. I think we can solve this issue after VarInfo is merged.

@xukai92
Copy link
Member

xukai92 commented Mar 25, 2019

I'm aware of that. Just wonder if there is an off-the-shelf GitHub service like that.

@yebai
Copy link
Member Author

yebai commented Apr 19, 2019

Just wonder if there is an off-the-shelf GitHub service like that.

This feature will be added to TuringBenchmarks.

@yebai yebai transferred this issue from TuringLang/AdvancedHMC.jl Jul 10, 2019
@storopoli
Copy link
Member

Guys do you need help to transfer Travis stuff to GitHub Actions?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants