[Core][Distributed] Add process-checkpoint lifecycle hooks for communicators (starting with Flashinfer) (#46877)

Signed-off-by: Schwinn Saereesitthipitak <schwinns@nvidia.com>
This commit is contained in:
Schwinn Saereesitthipitak
2026-07-26 14:47:50 -04:00
committed by GitHub
parent b68d7ef262
commit b5b61c622c
10 changed files with 185 additions and 4 deletions
@@ -93,7 +93,7 @@ def _can_use_flashinfer(hidden_states: torch.Tensor, tp_size: int) -> tuple[bool
max_token_num=max_token_num,
hidden_dim=hidden_size,
dtype=hidden_states.dtype,
group=get_tp_group().device_group,
group=get_tp_group().cpu_group,
)
if workspace is None:
return False, 0