Skip to content

Add cuda_graph_mode argument to explicitly control graph-unsafe opera…

58a1633
Select commit
Loading
Failed to load commit list.
Sign in for the full log view
Open

[Megatron-FSDP] Add conditional param.grad dereferencing logic to support full-iteration (FWD-BWD) CUDA graphability. #4663

Add cuda_graph_mode argument to explicitly control graph-unsafe opera…
58a1633
Select commit
Loading
Failed to load commit list.
check-approval
succeeded May 12, 2026 in 24s