[Diffusion] Add native SymmMem Fast Ulysses transport (#6340)

Signed-off-by: roG0d <baonudesifeizhai@gmail.com>
This commit is contained in:
baonudesifeizhai authored and GitHub committed 2026-08-30 04:14:21 +00:00
1 parent 99a2ae1fc8
commit 8be601ed7f
17 files changed
+749 -6

No files matched your search

+3
View File
@@ -223,6 +223,9 @@ class DiffusionParallelConfig:
sequence shapes across the ring group.
"""
ulysses_a2a_permute: bool = False
"""Use fused permute-free all-to-all for eligible strict Ulysses exchanges."""
cfg_parallel_size: int = 1
"""Number of ranks used to execute guidance passes in parallel."""