Skip to content

Fixes for training models with bf16 + freshly initialized optimizer via load_module_only#4141

Merged
loadams merged 8 commits intodeepspeedai:masterfrom haileyschoelkopf:new-fixJan 18, 2024

Commits

Commits on Aug 9, 2023

Commits on Aug 13, 2023

Commits on Nov 7, 2023

Commits on Dec 13, 2023

Commits on Jan 15, 2024

Commits on Jan 17, 2024