feat: update other MoE models deps (#2156)

python/sglang/srt/models/dbrx.py

@@ -24,7 +24,6 @@ from vllm.distributed import (
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_reduce,
 )
-from vllm.model_executor.layers.fused_moe import fused_moe
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.transformers_utils.configs.dbrx import DbrxConfig
@@ -37,6 +36,7 @@ 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 fused_moe
 from sglang.srt.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE,
     ParallelLMHead,

python/sglang/srt/models/deepseek.py

@@ -26,7 +26,6 @@ from vllm.distributed import (
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_reduce,
 )
-from vllm.model_executor.layers.fused_moe import fused_moe
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 
@@ -41,6 +40,7 @@ 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 fused_moe
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,

python/sglang/srt/models/mixtral.py

@@ -22,7 +22,6 @@ import torch
 from torch import nn
 from transformers import MixtralConfig
 from vllm.distributed import get_tensor_model_parallel_world_size
-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
 
@@ -36,6 +35,7 @@ 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.torchao_utils import apply_torchao_config_
+from sglang.srt.layers.triton_fused_moe import FusedMoE
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,

python/sglang/srt/models/olmoe.py

@@ -27,7 +27,6 @@ from vllm.distributed import (
     get_tensor_model_parallel_world_size,
     tensor_model_parallel_all_reduce,
 )
-from vllm.model_executor.layers.fused_moe import FusedMoE
 from vllm.model_executor.layers.linear import (
     MergedColumnParallelLinear,
     QKVParallelLinear,
@@ -43,6 +42,7 @@ from sglang.srt.layers.layernorm import RMSNorm
 from sglang.srt.layers.logits_processor import LogitsProcessor, LogitsProcessorOutput
 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,

python/sglang/srt/models/qwen2_moe.py

@@ -26,7 +26,6 @@ from vllm.distributed import (
     get_tensor_model_parallel_world_size,
     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.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.torchao_utils import apply_torchao_config_
+from sglang.srt.layers.triton_fused_moe import FusedMoE
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,

python/sglang/srt/models/xverse_moe.py

@@ -24,7 +24,6 @@ from vllm.distributed import (
     tensor_model_parallel_all_reduce,
 )
 from vllm.model_executor.layers.activation import SiluAndMul
-from vllm.model_executor.layers.fused_moe import fused_moe
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
     MergedColumnParallelLinear,
@@ -38,6 +37,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 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 fused_moe
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,