mirror of
https://github.com/RYDE-WORK/ktransformers.git
synced 2026-02-03 13:03:35 +08:00
commit
43e8848ddd
@ -727,7 +727,7 @@ class KDeepseekV2MoE(BaseInjectedModule, DeepseekV2MoE):
|
|||||||
)
|
)
|
||||||
return final_out
|
return final_out
|
||||||
|
|
||||||
class KMisrtalSparseMoEBlock(BaseInjectedModule, MixtralSparseMoeBlock):
|
class KMistralSparseMoEBlock(BaseInjectedModule, MixtralSparseMoeBlock):
|
||||||
|
|
||||||
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
|
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
|
||||||
""" """
|
""" """
|
||||||
|
|||||||
@ -19,7 +19,7 @@
|
|||||||
name: "^model\\.layers\\..*\\.block_sparse_moe$"
|
name: "^model\\.layers\\..*\\.block_sparse_moe$"
|
||||||
class: ktransformers.models.modeling_mixtral.MixtralSparseMoeBlock
|
class: ktransformers.models.modeling_mixtral.MixtralSparseMoeBlock
|
||||||
replace:
|
replace:
|
||||||
class: ktransformers.operators.experts.KMisrtalSparseMoEBlock
|
class: ktransformers.operators.experts.KMistralSparseMoEBlock
|
||||||
- match:
|
- match:
|
||||||
name: "^model\\.layers\\..*\\.block_sparse_moe\\.experts$"
|
name: "^model\\.layers\\..*\\.block_sparse_moe\\.experts$"
|
||||||
replace:
|
replace:
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user