diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 8f07d9fa..f7c597b2 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -37,7 +37,7 @@ jobs:
         pip install ${{ matrix.torch-spec }}
         ./.github/scripts/install_triton.sh
         pip install -r requirements.txt
-        python -m unittest discover -s test/ -p "*.py" -v
+        python -m unittest discover -s test/ -p "*.py" -v -t .
   test_cuda126_py312_a10g:
     name: test-cuda12.6-py3.12-a10g
     strategy:
@@ -63,4 +63,4 @@ jobs:
         pip install ${{ matrix.torch-spec }}
         ./.github/scripts/install_triton.sh
         pip install -r requirements.txt
-        python -m unittest discover -s test/ -p "*.py" -v
+        python -m unittest discover -s test/ -p "*.py" -v -t .
diff --git a/test/__init__.py b/test/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/test/test_closures.py b/test/test_closures.py
index 235bf825..e59a4897 100644
--- a/test/test_closures.py
+++ b/test/test_closures.py
@@ -98,7 +98,7 @@ def fn_with_global(x, tile) -> torch.Tensor:
 import triton.language as tl
 from torch._inductor.runtime.triton_helpers import math as tl_math
 
-import test_closures as _source_module
+import test.test_closures as _source_module
 
 @triton.jit
 def _sin_func_arg_kernel(a, _source_module_attr_global_tensor, out, a_size_0, _source_module_attr_global_tensor_stride_0, a_stride_0, out_stride_0, _BLOCK_SIZE_0: tl.constexpr):
@@ -280,7 +280,7 @@ def call_func_arg_on_host(a, alloc) -> torch.Tensor:
 import triton.language as tl
 from torch._inductor.runtime.triton_helpers import math as tl_math
 
-import test_closures as _source_module
+import test.test_closures as _source_module
 
 @triton.jit
 def _call_func_arg_on_host_kernel(a, out, a_size_0, a_stride_0, out_stride_0, _BLOCK_SIZE_0: tl.constexpr):
diff --git a/test/test_examples.py b/test/test_examples.py
index 25713732..69b792f9 100644
--- a/test/test_examples.py
+++ b/test/test_examples.py
@@ -241,7 +241,7 @@ def test_template_via_closure0(self):
 import triton.language as tl
 from torch._inductor.runtime import triton_helpers
 
-import test_examples as _global_source0
+import test.test_examples as _global_source0
 
 @triton.jit
 def _matmul_with_epilogue_kernel(x, y, epilogue_closure_0, out, _BLOCK_SIZE_0: tl.constexpr, _BLOCK_SIZE_1: tl.constexpr, _BLOCK_SIZE_2: tl.constexpr):
@@ -323,7 +323,7 @@ def test_template_via_closure1(self):
 import triton.language as tl
 from torch._inductor.runtime import triton_helpers
 
-import test_examples as _global_source0
+import test.test_examples as _global_source0
 
 @triton.jit
 def _matmul_with_epilogue_kernel(x, y, epilogue_closure_0, out, _BLOCK_SIZE_0: tl.constexpr, _BLOCK_SIZE_1: tl.constexpr, _BLOCK_SIZE_2: tl.constexpr):
@@ -401,7 +401,7 @@ def test_template_via_closure2(self):
 import triton.language as tl
 from torch._inductor.runtime import triton_helpers
 
-import test_examples as _global_source0
+import test.test_examples as _global_source0
 
 @triton.jit
 def _matmul_with_epilogue_kernel(x, y, out, _BLOCK_SIZE_0: tl.constexpr, _BLOCK_SIZE_1: tl.constexpr, _BLOCK_SIZE_2: tl.constexpr):