Use log_loss as tuning objective #197
Open
Conversation
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
flinder
requested review from
Lorenzo-Perini,
TaXxER and
danielha1
as code owners
meta-cla
bot
added
the
CLA Signed
flinder
force-pushed
the
export-D91884587
branch
from
5b66473 to
acd5efe
Compare
flinder
added a commit
to flinder/MCGrad-1
that referenced
this pull request
Jan 30, 2026
Summary: Pull Request resolved: facebookincubator#197 Use log_loss from sklearn instead of normalized_entropy as the tuning objective. This matches MCGrad's early stopping metric, providing directly comparable outputs between tuning and model training. The log_loss metric is a standard choice for probabilistic predictions and is widely understood. Differential Revision: D91884587
flinder
force-pushed
the
export-D91884587
branch
2 times, most recently
from
0c9ea67 to
80f26fd
Compare
flinder
added a commit
to flinder/MCGrad-1
that referenced
this pull request
Jan 30, 2026
Summary: Use log_loss from sklearn instead of normalized_entropy as the tuning objective. This matches MCGrad's early stopping metric, providing directly comparable outputs between tuning and model training. The log_loss metric is a standard choice for probabilistic predictions and is widely understood. Reviewed By: Lorenzo-Perini Differential Revision: D91884587
Summary: Add a `use_model_predictions` parameter to `tune_mcgrad_params` with `False` as the default. When False, returns the actual observed best trial parameters. When True, uses the Bayesian surrogate model's predicted best parameters. The default is False because with few tuning trials, the surrogate model may not be well-calibrated and could return suboptimal parameters that don't match the actual best observed trial. Users who want the model-predicted best can set this to True. Differential Revision: D91889101
Summary: Use log_loss from sklearn instead of normalized_entropy as the tuning objective. This matches MCGrad's early stopping metric, providing directly comparable outputs between tuning and model training. The log_loss metric is a standard choice for probabilistic predictions and is widely understood. Reviewed By: Lorenzo-Perini Differential Revision: D91884587
flinder
force-pushed
the
export-D91884587
branch
from
80f26fd to
b5ce578
Compare
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #197 +/- ##
=======================================
Coverage ? 94.92%
=======================================
Files ? 9
Lines ? 1753
Branches ? 0
=======================================
Hits ? 1664
Misses ? 89
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Summary:
Use log_loss from sklearn instead of normalized_entropy as the tuning objective.
This matches MCGrad's early stopping metric, providing directly comparable outputs
between tuning and model training. The log_loss metric is a standard choice for
probabilistic predictions and is widely understood.
Differential Revision: D91884587