A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 07:30 UTC
[BugFix][LoRA] TritonExperts is ModularMoEPath for FP8 models (#33393)
Signed-off-by: Danielle Robinson <dmmaddix@amazon.com> Co-authored-by: Danielle Robinson <dmmaddix@amazon.com>
Danielle Robinson committed
Jan 30, 2026 at 07:27 UTC
74898a70159f4217f1abf2a738c5cc6c4695cc44
1 file changed
+1
-3
vllm/lora/layers/fused_moe.py
+1
-3
@@ -143,9 +143,7 @@ class FusedMoEWithLoRA(BaseLayerWithLoRA):
143
m_fused_moe_fn.fused_experts, (MarlinExperts, UnfusedOAITritonExperts)
144
)
145
else:
146
- assert isinstance(
147
- m_fused_moe_fn.fused_experts, (MarlinExperts, TritonExperts)
148
- )
146
+ assert isinstance(m_fused_moe_fn.fused_experts, TritonExperts)
147
148
def fwd_decorator(layer, func):
149
def wrapper(*args, **kwargs):