feat: remove the dependency on FusedMoE (#2153)

python/sglang/srt/models/deepseek_v2.py

@@ -27,7 +27,6 @@ from vllm.distributed import (
     get_tp_group,
     tensor_model_parallel_all_reduce,
 )
-from vllm.model_executor.layers.fused_moe import FusedMoE
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
@@ -42,6 +41,7 @@ from sglang.srt.layers.linear import (
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.layers.radix_attention import RadixAttention
+from sglang.srt.layers.triton_fused_moe import FusedMoE
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,