feat(pipeline): add channel-ablation workflow and feature_channels config - #361
Merged
Merged
Conversation
jayhesselberth
added this pull request to stack #362
September 25, 2026 21:18
jayhesselberth
force-pushed
the
channel-ablation-pipeline
branch
from
September 25, 2026 21:24
11ba9d3 to
ee6c853
Compare
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #359; merge that first and this retargets to main. Measures what each feature channel contributes without re-preparing a corpus per arm: the train, validation and test corpora are prepared once per pair with
--feature-channels all, then each arm is oneleech model train --feature-channels ...run per seed on that shared corpus, scored byleech eval test. Arms are the default twelve as baseline, leave-one-out over each of the twelve, andlevel_skew/level_kurtosisadded singly and together, with free-form custom arms and three seeds by default. One compressed table reports each arm's mean, sd and delta vs baseline. Off by default (use_channel_ablation, targetall_channel_ablation). Also threads afeature_channelsconfig key into every rule that runsdata prepare; unset keeps today's behaviour.Verified by Snakemake dry run (107 jobs: 16 arms x 3 seeds) and
tests/test_pipeline_rules.py; not yet run on real data.https://claude.ai/code/session_01JDZiRc4dgzaA7nYFijxGbz