[CPU] Explicitly import sgl_kernel in CPU kernel tests (#35119)

test/registered/cpu/test_activation.py

@@ -1,6 +1,7 @@
 import sys
 
 import pytest
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler

test/registered/cpu/test_bmm.py

@@ -2,6 +2,7 @@ import itertools
 import unittest
 
 # TODO: use interface in cpu.py
+import sgl_kernel  # noqa: F401
 import torch
 import torch.nn as nn
 

test/registered/cpu/test_comm.py

@@ -5,6 +5,7 @@ import traceback
 import unittest
 from multiprocessing import Process
 
+import sgl_kernel  # noqa: F401
 import torch
 import torch.distributed as dist
 import torch.multiprocessing as mp

test/registered/cpu/test_decode.py

@@ -1,5 +1,6 @@
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 

test/registered/cpu/test_extend.py

@@ -1,5 +1,6 @@
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 

test/registered/cpu/test_gemm.py

@@ -1,6 +1,7 @@
 import unittest
 
 # TODO: use interface in cpu.py
+import sgl_kernel  # noqa: F401
 import torch
 import torch.nn as nn
 

test/registered/cpu/test_mamba.py

@@ -1,6 +1,7 @@
 import sys
 
 import pytest
+import sgl_kernel  # noqa: F401
 import torch
 import torch.nn.functional as F
 from torch.nn.functional import softplus

test/registered/cpu/test_mla.py

@@ -1,5 +1,6 @@
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 

test/registered/cpu/test_moe.py

@@ -3,6 +3,7 @@ import math
 import pytest
 
 # TODO: use interface in cpu.py
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.layers.amx_utils import CPUQuantMethod

test/registered/cpu/test_norm.py

@@ -2,6 +2,7 @@ import sys
 from typing import Optional, Tuple, Union
 
 import pytest
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.test.ci.ci_register import register_cpu_ci

test/registered/cpu/test_qkv_proj_with_rope.py

@@ -1,5 +1,6 @@
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.layers.quantization.fp8_utils import input_to_float8

test/registered/cpu/test_qwen3.py

@@ -1,6 +1,7 @@
 import sys
 
 import pytest
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.utils import is_host_cpu_arm64

test/registered/cpu/test_rope.py

@@ -1,5 +1,6 @@
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.layers.rotary_embedding import (

test/registered/cpu/test_shared_expert.py

@@ -2,6 +2,7 @@ import itertools
 import math
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.test.ci.ci_register import register_cpu_ci

test/registered/cpu/test_store_cache.py

@@ -1,6 +1,7 @@
 import sys
 
 import pytest
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.test.ci.ci_register import register_cpu_ci

test/registered/cpu/test_topk.py

@@ -1,6 +1,7 @@
 import itertools
 import unittest
 
+import sgl_kernel  # noqa: F401
 import torch
 
 from sglang.srt.eplb.expert_location_dispatch import ExpertLocationDispatchInfo