[MoE Refactor] FusedMoE/MoERunner inversion refactor (#41184)
Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
@@ -20,6 +20,7 @@ from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.attention import Attention
|
||||
from vllm.model_executor.layers.fused_moe import (
|
||||
FusedMoE,
|
||||
MoERunner,
|
||||
fused_moe_make_expert_params_mapping,
|
||||
)
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
@@ -640,7 +641,7 @@ class AfmoeForCausalLM(nn.Module, SupportsPP, SupportsEagle3, SupportsLoRA):
|
||||
self.num_moe_layers = config.num_hidden_layers - config.num_dense_layers
|
||||
self.num_expert_groups = config.n_group
|
||||
|
||||
self.moe_layers: list[FusedMoE] = []
|
||||
self.moe_layers: list[MoERunner] = []
|
||||
example_moe = None
|
||||
for layer in self.model.layers:
|
||||
if isinstance(layer, PPMissingLayer):
|
||||
|
||||
Reference in New Issue
Block a user