diff --git a/vllm/model_executor/__init__.py b/vllm/model_executor/__init__.py
index 76523845063ed..fb98f4a6b46f4 100644
--- a/vllm/model_executor/__init__.py
+++ b/vllm/model_executor/__init__.py
@@ -1,10 +1,7 @@
-from vllm.model_executor.layers.fused_moe.fused_moe import (
-    fused_moe, get_config_file_name, invoke_fused_moe_kernel,
-    moe_align_block_size)
+from vllm.model_executor.sampling_metadata import SamplingMetadata
+from vllm.model_executor.utils import set_random_seed
 
 __all__ = [
-    "fused_moe",
-    "get_config_file_name",
-    "moe_align_block_size",
-    "invoke_fused_moe_kernel",
+    "SamplingMetadata",
+    "set_random_seed",
 ]
diff --git a/vllm/model_executor/layers/fused_moe/__init__.py b/vllm/model_executor/layers/fused_moe/__init__.py
old mode 100644
new mode 100755
index 2926c7d1c8a76..851ed919ae3b8
--- a/vllm/model_executor/layers/fused_moe/__init__.py
+++ b/vllm/model_executor/layers/fused_moe/__init__.py
@@ -1,9 +1,8 @@
 from vllm.model_executor.layers.fused_moe.fused_moe import (
-    fused_experts, fused_moe, fused_topk, get_config_file_name)
+    fused_experts, fused_moe, fused_topk, get_config_file_name,
+    invoke_fused_moe_kernel, moe_align_block_size)
 
 __all__ = [
-    "fused_moe",
-    "fused_topk",
-    "fused_experts",
-    "get_config_file_name",
+    "fused_moe", "fused_topk", "fused_experts", "get_config_file_name",
+    "invoke_fused_moe_kernel", "moe_align_block_size"
 ]