diff --git a/.github/workflows/integration-tests.yml b/.github/workflows/integration-tests.yml
index 74ae590c4..544bce95b 100644
--- a/.github/workflows/integration-tests.yml
+++ b/.github/workflows/integration-tests.yml
@@ -24,6 +24,11 @@ jobs:
           cd python
           pip3 install -e .
 
+      - name: Unit tests
+        run: |
+          cd python/test/unit
+          pytest -vs .
+
       - name: Regression tests
         run: |
           cd python/test/regression
@@ -33,7 +38,3 @@ jobs:
           sudo nvidia-smi -i 0 -rgc
           sudo nvidia-smi -i 0 -rmc
 
-      - name: Unit tests
-        run: |
-          cd python/test/unit
-          pytest -vs .
\ No newline at end of file
diff --git a/python/test/regression/test_performance.py b/python/test/regression/test_performance.py
index a2c9f08ca..e205828d6 100644
--- a/python/test/regression/test_performance.py
+++ b/python/test/regression/test_performance.py
@@ -52,7 +52,7 @@ def test_matmul(M, N, K):
     cur_sm_clock = nvsmi(['clocks.current.sm'])[0]
     ref_sm_clock = 1350
     max_gpu_perf = 1e-6*80*8*128*cur_sm_clock
-    assert abs(cur_sm_clock - ref_sm_clock) < 5, f'GPU SMs must run at {ref_sm_clock} MHz'
+    assert abs(cur_sm_clock - ref_sm_clock) < 10, f'GPU SMs must run at {ref_sm_clock} MHz'
     a = torch.randn((M, K), dtype=torch.float16, device='cuda')
     b = torch.randn((K, N), dtype=torch.float16, device='cuda')
     fn = lambda: triton.ops.matmul(a, b)
@@ -95,7 +95,7 @@ def test_elementwise(N):
     cur_mem_clock = nvsmi(['clocks.current.memory'])[0]
     ref_mem_clock = 877
     max_gpu_perf = 512*2*ref_mem_clock*1e-3
-    assert abs(cur_mem_clock - ref_mem_clock) < 5, f'GPU memmory must run at {ref_mem_clock} MHz'
+    assert abs(cur_mem_clock - ref_mem_clock) < 10, f'GPU memmory must run at {ref_mem_clock} MHz'
     z = torch.empty((N, ), dtype=torch.float16, device='cuda')
     x = torch.randn_like(z)
     y = torch.randn_like(z)
diff --git a/python/triton/code_gen.py b/python/triton/code_gen.py
index 10561898f..2eec6248d 100644
--- a/python/triton/code_gen.py
+++ b/python/triton/code_gen.py
@@ -709,7 +709,7 @@ class JITFunction:
             return
         # create cache directory
         if not os.path.exists(cache_dir):
-            os.makedirs(cache_dir)
+            os.makedirs(cache_dir, exist_ok=True)
         # create md5 hash of src
         md5 = hashlib.md5()
         md5.update(self.src.encode('utf-8'))