[MoE] Move DEEP_GEMM into experts/ subdirectory (#39005)

Signed-off-by: Jackmin801 <ongjackm@gmail.com>
Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

tests/kernels/moe/modular_kernel_tools/mk_objects.py

@@ -7,14 +7,14 @@ import torch
 # Fused experts and PrepareFinalize imports
 import vllm.model_executor.layers.fused_moe.modular_kernel as mk
 from vllm.model_executor.layers.fused_moe import TritonExperts
-from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
-    BatchedDeepGemmExperts,
-)
 from vllm.model_executor.layers.fused_moe.config import (
     FusedMoEConfig,
     FusedMoEQuantConfig,
 )
-from vllm.model_executor.layers.fused_moe.deep_gemm_moe import DeepGemmExperts
+from vllm.model_executor.layers.fused_moe.experts.batched_deep_gemm_moe import (
+    BatchedDeepGemmExperts,
+)
+from vllm.model_executor.layers.fused_moe.experts.deep_gemm_moe import DeepGemmExperts
 from vllm.model_executor.layers.fused_moe.fused_batched_moe import (
     BatchedTritonExperts,
     NaiveBatchedExperts,

tests/kernels/moe/test_batched_deepgemm.py

@@ -5,10 +5,10 @@ import pytest
 import torch
 
 from vllm.model_executor.layers.fused_moe.activation import MoEActivation
-from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
+from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_config
+from vllm.model_executor.layers.fused_moe.experts.batched_deep_gemm_moe import (
     BatchedDeepGemmExperts,
 )
-from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_config
 from vllm.model_executor.layers.fused_moe.fused_batched_moe import (
     BatchedPrepareAndFinalize,
     BatchedTritonExperts,

tests/kernels/moe/test_block_fp8.py

@@ -28,7 +28,7 @@ from vllm.model_executor.layers.fused_moe.all2all_utils import (
 from vllm.model_executor.layers.fused_moe.config import (
     fp8_w8a8_moe_quant_config,
 )
-from vllm.model_executor.layers.fused_moe.deep_gemm_moe import (
+from vllm.model_executor.layers.fused_moe.experts.deep_gemm_moe import (
     _valid_deep_gemm_shape,
 )
 from vllm.model_executor.layers.fused_moe.triton_deep_gemm_moe import (

tests/kernels/moe/test_deepep_deepgemm_moe.py

@@ -47,10 +47,12 @@ if has_deep_ep():
     from .parallel_utils import DeepEPHTArgs, DeepEPLLArgs, make_deepep_a2a
 
 if has_deep_gemm():
-    from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
+    from vllm.model_executor.layers.fused_moe.experts.batched_deep_gemm_moe import (
         BatchedDeepGemmExperts,
     )
-    from vllm.model_executor.layers.fused_moe.deep_gemm_moe import DeepGemmExperts
+    from vllm.model_executor.layers.fused_moe.experts.deep_gemm_moe import (
+        DeepGemmExperts,
+    )
 
 requires_deep_ep = pytest.mark.skipif(
     not has_deep_ep(),

tests/kernels/moe/test_deepgemm.py

@@ -175,7 +175,7 @@ def test_deepgemm_vs_triton(m, n, k, topk, num_experts, monkeypatch, workspace_i
         mp.setenv("VLLM_USE_DEEP_GEMM", "1")
 
         _DeepGemmExperts = importlib.import_module(
-            "vllm.model_executor.layers.fused_moe.deep_gemm_moe"
+            "vllm.model_executor.layers.fused_moe.experts.deep_gemm_moe"
         ).DeepGemmExperts
 
         call_counter = {"cnt": 0}

tests/kernels/moe/test_silu_mul_fp8_quant_deep_gemm.py

@@ -7,7 +7,7 @@ import random
 import pytest
 import torch
 
-from vllm.model_executor.layers.fused_moe.batched_deep_gemm_moe import (
+from vllm.model_executor.layers.fused_moe.experts.batched_deep_gemm_moe import (
     persistent_masked_m_silu_mul_quant,
 )
 from vllm.model_executor.layers.quantization.utils.quant_utils import (