[Model] tp+ep support v1_loader (#5465)

* [Model] tp+ep support v1_loader

* fix

* fix mtp_linear

* fix mtp_linear

* fix

* fix

* fix v0 loader

* fix

* Add get_tensor for ep

* fix linear weight_loader

* fix typo

* fix
This commit is contained in:
Longzhi Wang
2025-12-18 14:31:54 +08:00
committed by GitHub
parent c89a62e550
commit d8587e987e
8 changed files with 48 additions and 20 deletions
@@ -86,6 +86,9 @@ class ParallelEHProjection(nn.Layer):
)
if self.tp_size > 1:
set_weight_attrs(self.linear.weight, {"output_dim": True})
if self.bias_key is not None:
set_weight_attrs(self.linear.bias, {"output_dim": True})
else:
self.linear = RowParallelLinear(
embedding_dim,