[Bugfix][MoE] Filter packed expert weights during EP loading (#49558)
Signed-off-by: aoshen02 <aoshen02@users.noreply.github.com> Signed-off-by: aoshen02 <aoshen@inferact.ai> Co-authored-by: aoshen02 <aoshen02@users.noreply.github.com>
This commit is contained in:
@@ -76,6 +76,6 @@ def should_skip_weight(
|
||||
# Only skip heavy weight tensors, never scale/metadata tensors.
|
||||
# Scale tensors are tiny and some backends need them from ALL experts
|
||||
# (e.g. FlashInfer NVFP4 computes a global max of activation scales).
|
||||
if not weight_name.endswith(".weight"):
|
||||
if not weight_name.endswith((".weight", ".weight_packed")):
|
||||
return False
|
||||
return eid not in local_expert_ids
|
||||
|
||||
Reference in New Issue
Block a user