diff --git a/.github/workflows/test_models.yml b/.github/workflows/test_models.yml
index 67c7054a..a46fc9fe 100644
--- a/.github/workflows/test_models.yml
+++ b/.github/workflows/test_models.yml
@@ -77,8 +77,8 @@ jobs:
           source turbine_venv/bin/activate
 
           pytest -v models/turbine_models/tests/sd_test.py
-          pytest -v models/turbine_models/tests/sdxl_test.py --device cpu --rt_device local-task --iree_target_triple x86_64-linux-gnu --num_inference_steps 2 -x
-          pytest -v models/turbine_models/tests/sdxl_test.py --device rocm --rt_device hip --iree_target_triple gfx90a --precision fp16 --attn_spec default -x
-          pytest -v models/turbine_models/tests/sdxl_test.py --device rocm --rt_device hip --iree_target_triple gfx90a --precision fp16 --clip_spec None --unet_spec None --vae_spec None --batch_size 2 -x
+          pytest -v models/turbine_models/tests/sdxl_test.py --device cpu --rt_device local-task --iree_target_triple x86_64-linux-gnu --num_inference_steps 2 -x -s
+          pytest -v models/turbine_models/tests/sdxl_test.py --device rocm --rt_device hip --iree_target_triple gfx90a --precision fp16 -x -s
+          pytest -v models/turbine_models/tests/sdxl_test.py --device rocm --rt_device hip --iree_target_triple gfx90a --precision fp16 --batch_size 2 -x
           pytest -v models/turbine_models/tests/sd3_test.py --device cpu --rt_device local-task --iree_target_triple x86_64-linux-gnu --num_inference_steps 2 -x
 
diff --git a/models/turbine_models/tests/sdxl_test.py b/models/turbine_models/tests/sdxl_test.py
index 6eba5b8e..3868f919 100644
--- a/models/turbine_models/tests/sdxl_test.py
+++ b/models/turbine_models/tests/sdxl_test.py
@@ -93,11 +93,7 @@ def test00_sdxl_pipe(self):
         decomp_attn = {
             "text_encoder": True,
             "unet": False,
-            "vae": (
-                False
-                if any(x in arguments["device"] for x in ["hip", "rocm"])
-                else True
-            ),
+            "vae": False,
         }
         self.pipe = SharkSDPipeline(
             arguments["hf_model_name"],
@@ -135,8 +131,12 @@ def test00_sdxl_pipe(self):
             True,  # return_img
         )
         assert output is not None
+        del output
+        del self.pipe
 
     def test01_sdxl_pipe_i8_punet(self):
+        if arguments["device"] not in ["rocm", "hip"]:
+            self.skipTest("Currently unimplemented/pending validation")
         from turbine_models.custom_models.sd_inference.sd_pipeline import (
             SharkSDPipeline,
         )
@@ -187,6 +187,8 @@ def test01_sdxl_pipe_i8_punet(self):
             True,  # return_img
         )
         assert output is not None
+        del output
+        del self.pipe
 
     def test02_PromptEncoder(self):
         if arguments["device"] in ["vulkan", "cuda"]: