update ci test (#1647)

Author: ovowei

kt-kernel/test/per_commit/test_moe_amx_accuracy_int4.py

@@ -99,7 +99,7 @@ def test_moe_amx_int4_accuracy():
         data=range(expert_num), device="cpu", dtype=torch.int64
     ).contiguous()
 
-    CPUInfer = kt_kernel_ext.CPUInfer(90)
+    CPUInfer = kt_kernel_ext.CPUInfer(60)
 
     with torch.inference_mode(mode=True):
         # Initialize MoE layers

kt-kernel/test/per_commit/test_moe_amx_accuracy_int4_1.py

@@ -99,7 +99,7 @@ def test_moe_amx_int4_1_accuracy():
         data=range(expert_num), device="cpu", dtype=torch.int64
     ).contiguous()
 
-    CPUInfer = kt_kernel_ext.CPUInfer(90)
+    CPUInfer = kt_kernel_ext.CPUInfer(60)
 
     with torch.inference_mode(mode=True):
         # Initialize MoE layers

kt-kernel/test/per_commit/test_moe_amx_accuracy_int4_1k.py

@@ -100,7 +100,7 @@ def test_moe_amx_int4_1k_accuracy():
         data=range(expert_num), device="cpu", dtype=torch.int64
     ).contiguous()
 
-    CPUInfer = kt_kernel_ext.CPUInfer(90)
+    CPUInfer = kt_kernel_ext.CPUInfer(60)
 
     with torch.inference_mode(mode=True):
         # Initialize MoE layers

kt-kernel/test/per_commit/test_moe_amx_accuracy_int8.py

@@ -99,7 +99,7 @@ def test_moe_amx_int8_accuracy():
         data=range(expert_num), device="cpu", dtype=torch.int64
     ).contiguous()
 
-    CPUInfer = kt_kernel_ext.CPUInfer(90)
+    CPUInfer = kt_kernel_ext.CPUInfer(60)
 
     with torch.inference_mode(mode=True):
         # Initialize MoE layers

kt-kernel/test/per_commit/test_moe_amx_bench_int4.py

@@ -45,9 +45,9 @@
 worker_config_dict = {
     "subpool_count": 2,
     "subpool_numa_map": [0, 1],
-    "subpool_thread_count": [45, 45],
+    "subpool_thread_count": [30, 30],
 }
-CPUINFER_PARAM = 90
+CPUINFER_PARAM = 60
 
 
 def get_git_commit():