Skip to content

Validate softmax temperature - #102

Open
sylvesterkaczmarek wants to merge 1 commit into
google-research:mainfrom
sylvesterkaczmarek:issue-93-softmax-temperature
Open

sylvesterkaczmarek wants to merge 1 commit into
google-research:mainfrom
sylvesterkaczmarek:issue-93-softmax-temperature

Conversation

@sylvesterkaczmarek

Copy link
Copy Markdown

Fixes #93.

Rejects zero and negative softmax temperatures before division so invalid configurations cannot silently invert predictions or produce NaN probabilities. Adds focused tests for non-positive values and the positive-temperature control case.

Validation:

  • uv run --extra dev python -m pytest -q tabfm/src/classifier_and_regressor_test.py -k SoftmaxTest
  • Python syntax compilation
  • git diff --check

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Negative softmax_temperature silently inverts predictions; zero yields all-nan probabilities

1 participant