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

Modify how to load prior distributions from edmf data #199

Merged
merged 1 commit into from
Jan 23, 2023
Merged

Conversation

szy21
Copy link
Member

@szy21 szy21 commented Jan 12, 2023

Purpose

This PR modifies how prior distributions are loaded to be consistent with recent changes in CalibrateEDMF. It also updates the EDMF data in ent-det-calibration.zip and ent-det-tked-tkee-stab-calibration.

Example output

exp_name = ent-det-calibration
image
exp_name = ent-det-tked-tkee-stab-calibration
image


  • I have read and checked the items on the review checklist.

@szy21 szy21 requested a review from odunbar January 12, 2023 01:07
Copy link
Collaborator

@odunbar odunbar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. Can we remove the manifest completely? I don't like having them hanging around. (Or is there a reason we had one in this example? I can't recall)
  2. Is the difference in the EDMF data just the way it's saved? Otherwise for reference could you post the resulting plot you get from the new data on your PR message. Thanks!

@szy21
Copy link
Member Author

szy21 commented Jan 12, 2023

  1. Can we remove the manifest completely? I don't like having them hanging around. (Or is there a reason we had one in this example? I can't recall)
  2. Is the difference in the EDMF data just the way it's saved? Otherwise for reference could you post the resulting plot you get from the new data on your PR message. Thanks!

I don't know why we had Manifest for this example, I have deleted it. The EDMF data has also changed (as I don't know exactly how Ignacio generated the data), and I have updated the commit message with the plot.

@charleskawczynski
Copy link
Member

(Or is there a reason we had one in this example? I can't recall)

Manifest files are useful for reproducibility

@szy21
Copy link
Member Author

szy21 commented Jan 12, 2023

(Or is there a reason we had one in this example? I can't recall)

Manifest files are useful for reproducibility

Ah, yes, I did have problems with jld2 compatibility when using a different version of Julia.

@odunbar
Copy link
Collaborator

odunbar commented Jan 12, 2023

Oh definitely, I get the general use for Manifests.

But we have had many more compatatbility issues trying to maintain them, when we had them fixed on GH, than reproducibility issues since we have removed them.

The JLD2 could also be sorted with a compat? Or by saving the data with a newer version of JLD2?

@codecov
Copy link

codecov bot commented Jan 12, 2023

Codecov Report

Base: 88.65% // Head: 88.65% // No change to project coverage 👍

Coverage data is based on head (2c35cc0) compared to base (b285e04).
Patch has no changes to coverable lines.

❗ Current head 2c35cc0 differs from pull request most recent head fd52016. Consider uploading reports for the commit fd52016 to get more accurate results

Additional details and impacted files
@@           Coverage Diff           @@
##             main     #199   +/-   ##
=======================================
  Coverage   88.65%   88.65%           
=======================================
  Files           4        4           
  Lines         388      388           
=======================================
  Hits          344      344           
  Misses         44       44           
Impacted Files Coverage Δ
src/Emulator.jl 93.63% <0.00%> (ø)
src/GaussianProcess.jl 93.00% <0.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

@odunbar
Copy link
Collaborator

odunbar commented Jan 12, 2023

@szy21 when you changed the JLD2 stuff, what did you do about the other dataset - the 5 dimensional one? Is it still compatible?

Edit - I just saw your PR comment. I would like to keep the 5D one or have a replacement please!

@szy21
Copy link
Member Author

szy21 commented Jan 15, 2023

@odunbar I updated the 5 parameter calibration data. The results are in the PR message. The parameters do not seem to be well constrained, but that's as expected I think.

Copy link
Collaborator

@odunbar odunbar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@szy21 Thanks!

As discussed, the example data is not ideal - however, CES seems to be working fine on it. To make plots smoother we can gather more samples:

  • Could you change (and ideally rerun) L193 to
    chain = MarkovChainMonteCarlo.sample(mcmc, 200_000; stepsize = new_step, discard_initial = 10_000)
  • Could you double (or measure) the approx run-time in the README.md

After this, merge is good!

Copy link
Collaborator

@odunbar odunbar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM! Thanks

@szy21
Copy link
Member Author

szy21 commented Jan 23, 2023

bors r+

@bors
Copy link
Contributor

bors bot commented Jan 23, 2023

@bors bors bot merged commit b8f0e96 into main Jan 23, 2023
@bors bors bot deleted the zs/edmf_data branch January 23, 2023 22:34
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

Successfully merging this pull request may close these issues.

3 participants