Skip to content

Add --discard-non-matching-values to remove non-matching x and scales when exporting into APPLgrid#400

Merged
Radonirinaunimi merged 9 commits intomasterfrom
fix-no-macthing-momentum
May 9, 2026
Merged

Add --discard-non-matching-values to remove non-matching x and scales when exporting into APPLgrid#400
Radonirinaunimi merged 9 commits intomasterfrom
fix-no-macthing-momentum

Conversation

@Radonirinaunimi
Copy link
Copy Markdown
Member

@Radonirinaunimi Radonirinaunimi commented May 5, 2026

TODO:

  • Add a CI test

@Radonirinaunimi Radonirinaunimi requested a review from cschwan May 5, 2026 20:22
Copy link
Copy Markdown
Contributor

@cschwan cschwan left a comment

Choose a reason for hiding this comment

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

Looks good to me, although we should refactor the code at some point, but probably not here and now.

I wonder whether we should merge the option into a single discard-non-matching-values that enables both. I don't think there's a reason to keep them separate. To make it backwards-compatible, we should add an alias (see also #291).

@Radonirinaunimi
Copy link
Copy Markdown
Member Author

I fully agree! And I just implemented it in 8ba5416.

@Radonirinaunimi Radonirinaunimi changed the title Add discard_non_matching_momentum to remove non-matching x when exporting into APPLgrid Add --discard-non-matching-values to remove non-matching x and scales when exporting into APPLgrid May 7, 2026
Copy link
Copy Markdown
Contributor

@cschwan cschwan left a comment

Choose a reason for hiding this comment

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

Can you extract the problematic subgrid(s)? The remaining subgrids are just dead weight.

@Radonirinaunimi
Copy link
Copy Markdown
Member Author

I am not exactly sure what you mean by this?

@cschwan
Copy link
Copy Markdown
Contributor

cschwan commented May 8, 2026

In general, we should keep the test-data size as small as possible, also to minimize the test run time. In this case I think we can just isolate the subgrid that's affected by the new flag and throw away the rest.

@cschwan
Copy link
Copy Markdown
Contributor

cschwan commented May 9, 2026

Commit f928b80 uses a smaller grid, which I created by stripping all log-orders and all bins except bin # 63. This bin has the most 'failing' values, including both factorization-scale and momentum-fraction types; the test grids goes down in size from 175 MB to 216 KB, which saves bandwidth and testing/CI time.

@cschwan cschwan added this to the v1.4 milestone May 9, 2026
@cschwan cschwan self-requested a review May 9, 2026 08:31
Copy link
Copy Markdown
Contributor

@cschwan cschwan left a comment

Choose a reason for hiding this comment

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

@Radonirinaunimi: have a look at my changes; as far as I am concerned this can be merged.

@Radonirinaunimi
Copy link
Copy Markdown
Member Author

Sorry @cschwan I was not able to look into it yesterday, so thanks for taking care of it. The changes look perfect to me.

@Radonirinaunimi Radonirinaunimi merged commit e5f11cc into master May 9, 2026
10 checks passed
@Radonirinaunimi Radonirinaunimi deleted the fix-no-macthing-momentum branch May 9, 2026 11:55
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.

2 participants