Skip to content

fix: satisfy assign_splits non-empty guarantee when all tasks hash to test #274

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.

Sign up for GitHub

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

Closed
RohithPariki wants to merge 1 commit into from
Closed

fix: satisfy assign_splits non-empty guarantee when all tasks hash to test #274

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 11 additions & 1 deletion skillopt_sleep/mine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -341,9 +341,19 @@ def _promote_one(*, to: str, from_splits: set[str]) -> None:
# Only promote from train so hash-assigned test tasks stay untouched.
if len(real) >= 2 and not any(t.split == "val" for t in real):
_promote_one(to="val", from_splits={"train"})
if not any(t.split == "val" for t in real):
import logging
logging.warning("holdout_leaked: test task promoted to val to satisfy non-empty guarantee")
_promote_one(to="val", from_splits={"test"})
# Guarantee a train pool exists when possible; never borrow from test.
if not any(t.split == "train" for t in tasks) and len(real) >= 2:
_promote_one(to="train", from_splits={"val"})
val_count = sum(1 for t in real if t.split == "val")
if val_count > 1:
_promote_one(to="train", from_splits={"val"})
else:
import logging
logging.warning("holdout_leaked: test task promoted to train to satisfy non-empty guarantee")
_promote_one(to="train", from_splits={"test"})
return tasks


Expand Down