Draft: Resolve "Wasserstein-GAN throwaway experiment (fast single-pass eval)" #14

Merged

lbogner merged 7 commits from 3-wasserstein-gan-throwaway-experiment-fast-single-pass-eval into master 2026-07-23 15:32:49 +02:00

Conversation 3 Commits 7 Files Changed 14

+1294 -105

lbogner commented 2026-07-22 10:35:37 +02:00 (Migrated from gitlab.etp.kit.edu)

Copy Link

Copy Source

Closes #3

Closes #3

lbogner commented 2026-07-22 10:41:58 +02:00 (Migrated from gitlab.etp.kit.edu)

Copy Link

Copy Source

added 3 commits

• a6bb142a...4d190696 - 3 commits from branch master

Compare with previous version

added 3 commits <ul><li>a6bb142a...4d190696 - 3 commits from branch <code>master</code></li></ul> [Compare with previous version](/lbogner/giant/-/merge_requests/6/diffs?diff_id=5622&start_sha=a6bb142a4086c067003335d8a14d1c854702ec6a)

lbogner commented 2026-07-22 11:21:49 +02:00 (Migrated from gitlab.etp.kit.edu)

Copy Link

Copy Source

added 1 commit

• 44b0a92e - Add WGAN-GP mode as a throwaway fast-eval experiment

Compare with previous version

added 1 commit <ul><li>44b0a92e - Add WGAN-GP mode as a throwaway fast-eval experiment</li></ul> [Compare with previous version](/lbogner/giant/-/merge_requests/6/diffs?diff_id=5623&start_sha=4d190696fecce5d71a6ed81d589629093b3462fd)

lbogner commented 2026-07-22 11:22:47 +02:00 (Migrated from gitlab.etp.kit.edu)

Copy Link

Copy Source

added 1 commit

• 2e5268f4 - Ignore the scratchpad working directory

Compare with previous version

added 1 commit <ul><li>2e5268f4 - Ignore the scratchpad working directory</li></ul> [Compare with previous version](/lbogner/giant/-/merge_requests/6/diffs?diff_id=5624&start_sha=44b0a92e6734cc1954111767bb0261a3e73075ed)

lars added 1 commit 2026-07-23 14:15:43 +02:00

Add router gating diagnostic for MoE checkpoints ... 0d2967fa6b

plot_router_gating visualizes soft expert gate weights vs. a continuous
routing axis (e.g. pre-step energy), binned into equal-population
quantiles and stacked to show the router's soft decision boundaries.
Wired into rollout_validation.ipynb as a new notebook-only section
that loads a checkpoint's Router directly, since gate weights aren't
present in rollout/predict parquet output.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

lars added 1 commit 2026-07-23 14:33:20 +02:00

Add Gitea Actions CI pipeline ... 4e77e3fc4b

Three-stage workflow: lint/format/type-check (ruff, ty), pytest,
and (master pushes only) a version-bump + uv build + publish to the
Gitea package registry. Bump commit is tagged [skip ci] to avoid
retriggering the workflow.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

lars added 1 commit 2026-07-23 14:35:57 +02:00

Quote the on: key in the CI workflow ... faac19edf3

Unquoted on: gets parsed as the YAML 1.1 boolean true instead of the
string "on", so Gitea couldn't find a trigger key and didn't register
the workflow at all.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

lars added 2 commits 2026-07-23 14:38:53 +02:00

Split CI lint stage into parallel jobs ... 9162ba7ae0

ruff check, ruff format check, ty check, and pytest now run as four
independent jobs instead of one sequential lint job gating test —
faster wall-clock CI since none of these checks depend on each other.
build still waits on all four before bumping/publishing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Reformat rollout_validation.ipynb with ruff format ... 8d46f27d1b

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

lars closed this pull request 2026-07-23 15:28:16 +02:00

lars reopened this pull request 2026-07-23 15:31:52 +02:00

lars marked the pull request as ready for review 2026-07-23 15:32:07 +02:00

lars merged commit 4f785c43e6 into master 2026-07-23 15:32:49 +02:00

lars deleted branch 3-wasserstein-gan-throwaway-experiment-fast-single-pass-eval 2026-07-23 15:32:49 +02:00

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

Clear labels

architecture

experiment

investigation

performance

phase2

validation

No Label architecture experiment

Milestone

No items

No Milestone

Projects

Clear projects

No project

Assignees

Clear assignees

lars (Lars Bogner)

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: lars/giant#14