Rewrite the README, and untrack the FAERS build cache - #4
Merged
Conversation
added 2 commits
August 10, 2026 15:38
460 MB of downloaded quarterly archives and parsed tallies reached main. The .gitignore rule covering them was written on the pv_signal_triage branch, so a 'git add -A' on a branch cut from main before that swept them in, and the squash merge carried them. This removes them from tracking. It does not remove them from history, so a clone still pays for them until the history is rewritten, which needs a force-push to main and is therefore left as a decision rather than taken. The cache is a build artifact. scripts/build_pv_snapshot.py re-downloads and re-derives it, and the snapshot it produces is digest-pinned, so nothing about reproducibility depends on shipping it.
The README described one environment and predated the harness features the last two changes added. It also read like a specification when the first thing a reader needs is to understand what the repo is for. Content: covers safety_judgment, pv_signal_triage and target_triage, each with its baseline table; documents cost-weighted mistakes and the listing floor alongside the existing probes; fixes the environment count badge; credits FAERS; points the quickstart at safety_judgment. Language: shorter sentences, plainer words, fewer clauses per thought. The substance and the caveats are unchanged, including that the corpus is single-author, that pv_signal_triage's thresholds were raised after seeing results, and that its gap is much narrower than safety_judgment's. Also marks cost_weighted_negatives in safety_judgment's manifest, which was using severity-priced false positives without declaring it.
aaygan29
added a commit
that referenced
this pull request
Aug 10, 2026
* Untrack the FAERS build cache 460 MB of downloaded quarterly archives and parsed tallies reached main. The .gitignore rule covering them was written on the pv_signal_triage branch, so a 'git add -A' on a branch cut from main before that swept them in, and the squash merge carried them. This removes them from tracking. It does not remove them from history, so a clone still pays for them until the history is rewritten, which needs a force-push to main and is therefore left as a decision rather than taken. The cache is a build artifact. scripts/build_pv_snapshot.py re-downloads and re-derives it, and the snapshot it produces is digest-pinned, so nothing about reproducibility depends on shipping it. * Rewrite the README: simpler language, all three environments The README described one environment and predated the harness features the last two changes added. It also read like a specification when the first thing a reader needs is to understand what the repo is for. Content: covers safety_judgment, pv_signal_triage and target_triage, each with its baseline table; documents cost-weighted mistakes and the listing floor alongside the existing probes; fixes the environment count badge; credits FAERS; points the quickstart at safety_judgment. Language: shorter sentences, plainer words, fewer clauses per thought. The substance and the caveats are unchanged, including that the corpus is single-author, that pv_signal_triage's thresholds were raised after seeing results, and that its gap is much narrower than safety_judgment's. Also marks cost_weighted_negatives in safety_judgment's manifest, which was using severity-priced false positives without declaring it. --------- Co-authored-by: Aayush Gandhi <aayushgandhi@Aayushs-MacBook-Air.local>
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.
README
The README described one environment and predated the harness features the last two changes
added. It also read like a specification, when the first thing a reader needs is to understand
what the repo is for.
Content brought current:
safety_judgmentLanguage simplified: shorter sentences, plainer words, fewer clauses per thought.
The substance and every caveat are unchanged: the
safety_judgmentcorpus is single-authorand needs multi-rater review,
pv_signal_triage's thresholds were raised after seeing results,its gap is much narrower than
safety_judgment's, andtarget_triageis not validated as ameasure of drug discovery judgement.
Also declares
cost_weighted_negativesinsafety_judgment's manifest, which had been usingseverity-priced false positives without saying so.
Untracking the build cache
This does not fully fix the problem, and the remaining part needs a decision.
460 MB of downloaded FAERS quarterly archives and parsed tallies reached
main. The.gitignorerule covering them was written on thepv_signal_triagebranch, so agit add -Aon a branch cut from
mainbefore that swept them in, and the squash merge carried them.This commit removes them from tracking. It does not remove them from history, so a fresh
clone still downloads roughly 460 MB it has no use for. Purging that needs a history rewrite
and a force-push to
main, which is not something to do unilaterally on a public repository.Nothing about reproducibility depends on shipping the cache. It is a build artifact;
scripts/build_pv_snapshot.pyre-downloads and re-derives it, and the snapshot it produces isdigest-pinned independently.