Skip to content

Add separate final normalization for Multi-Token Prediction in DeepSeek models. - #5142

Open
copybara-service[bot] wants to merge 1 commit into
mainfrom
test_975503733
Open

Add separate final normalization for Multi-Token Prediction in DeepSeek models.#5142
copybara-service[bot] wants to merge 1 commit into
mainfrom
test_975503733

Commits

Commits on Sep 4, 2026