diff --git a/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py b/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py
index 9ee84643c9..ecc7fc57e9 100644
--- a/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py
+++ b/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py
@@ -10,7 +10,7 @@
 from torch.fx.node import Argument, Node, Target
 from torch_tensorrt import ENABLED_FEATURES
 from torch_tensorrt._features import needs_not_tensorrt_rtx
-from torch_tensorrt._utils import is_tensorrt_version_supported, is_thor
+from torch_tensorrt._utils import is_tensorrt_version_supported
 from torch_tensorrt.dynamo._settings import CompilationSettings
 from torch_tensorrt.dynamo._SourceIR import SourceIR
 from torch_tensorrt.dynamo.conversion import impl
@@ -429,7 +429,7 @@ def index_nonbool_validator(
     node: Node, settings: Optional[CompilationSettings] = None
 ) -> bool:
     # for thor and tensorrt_rtx, we don't support boolean indices, due to nonzero op not supported
-    if is_thor() or ENABLED_FEATURES.tensorrt_rtx:
+    if ENABLED_FEATURES.tensorrt_rtx:
         index = node.args[1]
         for ind in index:
             if ind is not None:
@@ -3621,18 +3621,10 @@ def aten_ops_full(
     )
 
 
-def nonzero_validator(
-    node: Node, settings: Optional[CompilationSettings] = None
-) -> bool:
-    return not is_thor()
-
-
 # currently nonzero is not supported for tensorrt_rtx
 # TODO: lan to add the nonzero support once tensorrt_rtx team has added the support
-# TODO: apbose to remove the capability validator once thor bug resolve in NGC
 @dynamo_tensorrt_converter(
     torch.ops.aten.nonzero.default,
-    capability_validator=nonzero_validator,
     supports_dynamic_shapes=True,
     requires_output_allocator=True,
 )
diff --git a/tests/py/dynamo/conversion/test_arange_aten.py b/tests/py/dynamo/conversion/test_arange_aten.py
index 968611a4ec..7705590e2e 100644
--- a/tests/py/dynamo/conversion/test_arange_aten.py
+++ b/tests/py/dynamo/conversion/test_arange_aten.py
@@ -5,15 +5,10 @@
 import torch_tensorrt
 from parameterized import parameterized
 from torch.testing._internal.common_utils import run_tests
-from torch_tensorrt._utils import is_tegra_platform, is_thor
 
 from .harness import DispatchTestCase
 
 
-@unittest.skipIf(
-    is_thor() or is_tegra_platform(),
-    "Skipped on Thor and Tegra platforms",
-)
 class TestArangeConverter(DispatchTestCase):
     @parameterized.expand(
         [
diff --git a/tests/py/dynamo/conversion/test_cumsum_aten.py b/tests/py/dynamo/conversion/test_cumsum_aten.py
index a677212cb1..c405d0e2f0 100644
--- a/tests/py/dynamo/conversion/test_cumsum_aten.py
+++ b/tests/py/dynamo/conversion/test_cumsum_aten.py
@@ -5,15 +5,10 @@
 import torch_tensorrt
 from parameterized import parameterized
 from torch.testing._internal.common_utils import run_tests
-from torch_tensorrt._utils import is_tegra_platform, is_thor
 
 from .harness import DispatchTestCase
 
 
-@unittest.skipIf(
-    is_thor() or is_tegra_platform(),
-    "Skipped on Thor and Tegra platforms",
-)
 class TestCumsumConverter(DispatchTestCase):
     @parameterized.expand(
         [
diff --git a/tests/py/dynamo/conversion/test_index_aten.py b/tests/py/dynamo/conversion/test_index_aten.py
index e34dc48dd5..6ed5f911f2 100644
--- a/tests/py/dynamo/conversion/test_index_aten.py
+++ b/tests/py/dynamo/conversion/test_index_aten.py
@@ -6,7 +6,6 @@
 from parameterized import parameterized
 from torch.testing._internal.common_utils import run_tests
 from torch_tensorrt import ENABLED_FEATURES, Input
-from torch_tensorrt._utils import is_tegra_platform, is_thor
 
 from .harness import DispatchTestCase
 
@@ -114,8 +113,8 @@ def forward(self, input):
         ]
     )
     @unittest.skipIf(
-        is_thor() or ENABLED_FEATURES.tensorrt_rtx,
-        "Skipped on Thor or tensorrt_rtx due to nonzero not supported",
+        ENABLED_FEATURES.tensorrt_rtx,
+        "Skipped on tensorrt_rtx due to nonzero not supported",
     )
     def test_index_constant_bool_mask(self, _, index, input):
         class TestModule(torch.nn.Module):
@@ -149,8 +148,8 @@ def forward(self, x, index0):
         )
 
     @unittest.skipIf(
-        is_thor() or ENABLED_FEATURES.tensorrt_rtx,
-        "Skipped on Thor or tensorrt_rtx due to nonzero not supported",
+        ENABLED_FEATURES.tensorrt_rtx,
+        "Skipped on tensorrt_rtx due to nonzero not supported",
     )
     def test_index_zero_two_dim_ITensor_mask(self):
         class TestModule(nn.Module):
@@ -163,10 +162,6 @@ def forward(self, x, index0):
         index0 = torch.tensor([True, False])
         self.run_test(TestModule(), [input, index0], enable_passes=True)
 
-    @unittest.skipIf(
-        is_thor(),
-        "Skipped on Thor due to nonzero not supported",
-    )
     def test_index_zero_index_three_dim_ITensor(self):
         class TestModule(nn.Module):
             def forward(self, x, index0):
@@ -180,8 +175,8 @@ def forward(self, x, index0):
         self.run_test(TestModule(), [input, index0])
 
     @unittest.skipIf(
-        is_thor() or ENABLED_FEATURES.tensorrt_rtx,
-        "Skipped on Thor or tensorrt_rtx due to nonzero not supported",
+        ENABLED_FEATURES.tensorrt_rtx,
+        "Skipped on tensorrt_rtx due to nonzero not supported",
     )
     def test_index_zero_index_three_dim_mask_ITensor(self):
         class TestModule(nn.Module):
@@ -252,7 +247,7 @@ def forward(self, input):
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "nonzero is not supported for tensorrt_rtx",
 )
 class TestIndexDynamicInputNonDynamicIndexConverter(DispatchTestCase):
diff --git a/tests/py/dynamo/conversion/test_nonzero_aten.py b/tests/py/dynamo/conversion/test_nonzero_aten.py
index 641cc7c098..b81644ed54 100644
--- a/tests/py/dynamo/conversion/test_nonzero_aten.py
+++ b/tests/py/dynamo/conversion/test_nonzero_aten.py
@@ -6,13 +6,12 @@
 from parameterized import parameterized
 from torch.testing._internal.common_utils import run_tests
 from torch_tensorrt import Input
-from torch_tensorrt._utils import is_tegra_platform, is_thor
 
 from .harness import DispatchTestCase
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "nonzero is not supported for tensorrt_rtx",
 )
 class TestNonZeroConverter(DispatchTestCase):
diff --git a/tests/py/dynamo/conversion/test_sym_size.py b/tests/py/dynamo/conversion/test_sym_size.py
index 770dd75504..968ecd322b 100644
--- a/tests/py/dynamo/conversion/test_sym_size.py
+++ b/tests/py/dynamo/conversion/test_sym_size.py
@@ -4,15 +4,10 @@
 import torch.nn as nn
 from parameterized import parameterized
 from torch.testing._internal.common_utils import run_tests
-from torch_tensorrt._utils import is_thor
 
 from .harness import DispatchTestCase
 
 
-@unittest.skipIf(
-    is_thor(),
-    "Skipped on Thor",
-)
 class TestSymSizeConverter(DispatchTestCase):
     @parameterized.expand(
         [
diff --git a/tests/py/dynamo/models/test_export_kwargs_serde.py b/tests/py/dynamo/models/test_export_kwargs_serde.py
index dabbad3cc8..44a2e0d6c7 100644
--- a/tests/py/dynamo/models/test_export_kwargs_serde.py
+++ b/tests/py/dynamo/models/test_export_kwargs_serde.py
@@ -1,6 +1,5 @@
 # type: ignore
 import os
-import tempfile
 import unittest
 
 import pytest
@@ -22,7 +21,7 @@
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_custom_model():
+def test_custom_model(tmpdir):
     class net(nn.Module):
         def __init__(self):
             super().__init__()
@@ -75,7 +74,7 @@ def forward(self, x, b=5, c=None, d=None):
     )
 
     # Save the module
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     torchtrt.save(trt_gm, trt_ep_path, retrace=False)
     # Clean up model env
     torch._dynamo.reset()
@@ -83,7 +82,7 @@ def forward(self, x, b=5, c=None, d=None):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_custom_model_with_dynamo_trace():
+def test_custom_model_with_dynamo_trace(tmpdir):
     class net(nn.Module):
         def __init__(self):
             super().__init__()
@@ -137,7 +136,7 @@ def forward(self, x, b=5, c=None, d=None):
     )
 
     # Save the module
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     torchtrt.save(trt_gm, trt_ep_path, retrace=False)
     # Clean up model env
     torch._dynamo.reset()
@@ -145,7 +144,7 @@ def forward(self, x, b=5, c=None, d=None):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_custom_model_with_dynamo_trace_dynamic():
+def test_custom_model_with_dynamo_trace_dynamic(tmpdir):
     class net(nn.Module):
         def __init__(self):
             super().__init__()
@@ -208,7 +207,7 @@ def forward(self, x, b=5, c=None, d=None):
     )
 
     # Save the module
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     torchtrt.save(trt_gm, trt_ep_path, retrace=False)
     # Clean up model env
     torch._dynamo.reset()
@@ -216,7 +215,7 @@ def forward(self, x, b=5, c=None, d=None):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_custom_model_with_dynamo_trace_kwarg_dynamic():
+def test_custom_model_with_dynamo_trace_kwarg_dynamic(tmpdir):
     ir = "dynamo"
 
     class net(nn.Module):
@@ -298,7 +297,7 @@ def forward(self, x, b=None, c=None, d=None, e=[]):
         msg=f"CustomKwargs Module TRT outputs don't match with the original model. Cosine sim score: {cos_sim} Threshold: {COSINE_THRESHOLD}",
     )
     # Save the module
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     torchtrt.save(trt_gm, trt_ep_path, retrace=False)
     # Clean up model env
     torch._dynamo.reset()
@@ -306,7 +305,7 @@ def forward(self, x, b=None, c=None, d=None, e=[]):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_custom_model_with_dynamo_trace_kwarg_dynamic():
+def test_custom_model_with_dynamo_trace_kwarg_dynamic(tmpdir):
     ir = "dynamo"
 
     class net(nn.Module):
@@ -388,7 +387,7 @@ def forward(self, x, b=None, c=None, d=None, e=[]):
         msg=f"CustomKwargs Module TRT outputs don't match with the original model. Cosine sim score: {cos_sim} Threshold: {COSINE_THRESHOLD}",
     )
     # Save the module
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     torchtrt.save(trt_gm, trt_ep_path, retrace=False)
     # Clean up model env
     torch._dynamo.reset()
diff --git a/tests/py/dynamo/models/test_export_serde.py b/tests/py/dynamo/models/test_export_serde.py
index c5b007e34b..3209177120 100644
--- a/tests/py/dynamo/models/test_export_serde.py
+++ b/tests/py/dynamo/models/test_export_serde.py
@@ -1,7 +1,6 @@
 import importlib
 import os
 import platform
-import tempfile
 import unittest
 
 import pytest
@@ -17,16 +16,15 @@
 if importlib.util.find_spec("torchvision"):
     import torchvision.models as models
 
-trt_ep_path = os.path.join(tempfile.gettempdir(), "trt.ep")
-
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_base_full_compile(ir):
+def test_base_full_compile(ir, tmpdir):
     """
     This tests export serde functionality on a base model
     which is fully TRT convertible
     """
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
 
     class MyModule(torch.nn.Module):
         def __init__(self):
@@ -76,11 +74,12 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_base_full_compile_multiple_outputs(ir):
+def test_base_full_compile_multiple_outputs(ir, tmpdir):
     """
     This tests export serde functionality on a base model
     with multiple outputs which is fully TRT convertible
     """
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
 
     class MyModule(torch.nn.Module):
         def __init__(self):
@@ -136,11 +135,12 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_no_compile(ir):
+def test_no_compile(ir, tmpdir):
     """
     This tests export serde functionality on a model
     which won't convert to TRT because of min_block_size=5 constraint
     """
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
 
     class MyModule(torch.nn.Module):
         def __init__(self):
@@ -195,13 +195,15 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_hybrid_relu_fallback(ir):
+def test_hybrid_relu_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a hybrid
     model with Pytorch and TRT segments. Relu (unweighted) layer is forced to
     fallback
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -258,10 +260,13 @@ def forward(self, x):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18(ir):
+def test_resnet18(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     model = models.resnet18().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -303,10 +308,12 @@ def test_resnet18(ir):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18_cpu_offload(ir):
+def test_resnet18_cpu_offload(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model
     """
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     model = models.resnet18().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -355,10 +362,13 @@ def test_resnet18_cpu_offload(ir):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18_dynamic(ir):
+def test_resnet18_dynamic(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     model = models.resnet18().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -395,10 +405,13 @@ def test_resnet18_dynamic(ir):
 @unittest.skipIf(
     not importlib.util.find_spec("torchvision"), "torchvision not installed"
 )
-def test_resnet18_torch_exec_ops_serde(ir):
+def test_resnet18_torch_exec_ops_serde(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     model = models.resnet18().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -426,12 +439,14 @@ def test_resnet18_torch_exec_ops_serde(ir):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_hybrid_conv_fallback(ir):
+def test_hybrid_conv_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a hybrid
     model where a conv (a weighted layer)  has been forced to fallback to Pytorch.
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -487,12 +502,14 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_hybrid_conv_fallback_cpu_offload(ir):
+def test_hybrid_conv_fallback_cpu_offload(ir, tmpdir):
     """
     This tests export save and load functionality on a hybrid
     model where a conv (a weighted layer)  has been forced to fallback to Pytorch.
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -549,13 +566,15 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_arange_export(ir):
+def test_arange_export(ir, tmpdir):
     """
     This tests export save and load functionality on a arange static graph
     Here the arange output is a static constant (which is registered as input to the graph)
     in the exporter.
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -607,7 +626,7 @@ def forward(self, x):
 
 
 @pytest.mark.unit
-def test_save_load_ts(ir):
+def test_save_load_ts(ir, tmpdir):
     """
     This tests save/load API on Torchscript format (model still compiled using dynamo workflow)
     """
@@ -624,6 +643,7 @@ def forward(self, x):
             mul = relu * 0.5
             return mul
 
+    ts_path = os.path.join(tmpdir, "trt.ts")
     model = MyModule().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -641,9 +661,9 @@ def forward(self, x):
     )
     outputs_trt = trt_gm(input)
     # Save it as torchscript representation
-    torchtrt.save(trt_gm, "./trt.ts", output_format="torchscript", inputs=[input])
+    torchtrt.save(trt_gm, ts_path, output_format="torchscript", inputs=[input])
 
-    trt_ts_module = torchtrt.load("./trt.ts")
+    trt_ts_module = torchtrt.load(ts_path)
     outputs_trt_deser = trt_ts_module(input)
 
     cos_sim = cosine_similarity(outputs_trt, outputs_trt_deser)
diff --git a/tests/py/dynamo/models/test_model_refit.py b/tests/py/dynamo/models/test_model_refit.py
index e6b7f6e2a4..813f2f848d 100644
--- a/tests/py/dynamo/models/test_model_refit.py
+++ b/tests/py/dynamo/models/test_model_refit.py
@@ -1,6 +1,5 @@
 import importlib
 import os
-import tempfile
 import unittest
 
 import pytest
@@ -531,8 +530,9 @@ def test_refit_one_engine_bert_with_weightmap():
     "Refit feature is not supported in Python 3.13 or higher",
 )
 @pytest.mark.unit
-def test_refit_one_engine_inline_runtime_with_weightmap():
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+def test_refit_one_engine_inline_runtime_with_weightmap(tmpdir):
+
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     model = models.resnet18(pretrained=False).eval().to("cuda")
     model2 = models.resnet18(pretrained=True).eval().to("cuda")
     inputs = [torch.randn((1, 3, 224, 224)).to("cuda")]
@@ -888,8 +888,8 @@ def test_refit_one_engine_bert_without_weightmap():
     "Refit feature is not supported in Python 3.13 or higher",
 )
 @pytest.mark.unit
-def test_refit_one_engine_inline_runtime_without_weightmap():
-    trt_ep_path = os.path.join(tempfile.gettempdir(), "compiled.ep")
+def test_refit_one_engine_inline_runtime_without_weightmap(tmpdir):
+    trt_ep_path = os.path.join(tmpdir, "compiled.ep")
     model = models.resnet18(pretrained=True).eval().to("cuda")
     model2 = models.resnet18(pretrained=False).eval().to("cuda")
     inputs = [torch.randn((1, 3, 224, 224)).to("cuda")]
diff --git a/tests/py/dynamo/models/test_reexport.py b/tests/py/dynamo/models/test_reexport.py
index 7c414def52..9636c9d91a 100644
--- a/tests/py/dynamo/models/test_reexport.py
+++ b/tests/py/dynamo/models/test_reexport.py
@@ -1,6 +1,5 @@
 import importlib
 import os
-import tempfile
 import unittest
 
 import pytest
@@ -14,12 +13,10 @@
 if importlib.util.find_spec("torchvision"):
     import torchvision.models as models
 
-trt_ep_path = os.path.join(tempfile.gettempdir(), "trt.ep")
-
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_base_full_compile(ir):
+def test_base_full_compile(ir, tmpdir):
     """
     This tests export serde functionality on a base model
     which is fully TRT convertible
@@ -56,6 +53,8 @@ def forward(self, x):
 
     # Reexport
     trt_exp_program = torch.export.export(trt_module, (input,), strict=False)
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     torch.export.save(trt_exp_program, trt_ep_path)
 
     deser_trt_module = torchtrt.load(trt_ep_path).module()
@@ -76,12 +75,14 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_base_full_compile_multiple_outputs(ir):
+def test_base_full_compile_multiple_outputs(ir, tmpdir):
     """
     This tests export serde functionality on a base model
     with multiple outputs which is fully TRT convertible
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -139,7 +140,7 @@ def forward(self, x):
 
 @pytest.mark.unit
 @pytest.mark.critical
-def test_no_compile(ir):
+def test_no_compile(ir, tmpdir):
     """
     This tests export serde functionality on a model
     which won't convert to TRT because of min_block_size=5 constraint
@@ -157,6 +158,7 @@ def forward(self, x):
             relu = self.relu(conv)
             return conv, relu
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
     model = MyModule().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -200,13 +202,15 @@ def forward(self, x):
 
 
 @pytest.mark.unit
-def test_hybrid_relu_fallback(ir):
+def test_hybrid_relu_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a hybrid
     model with Pytorch and TRT segments. Relu (unweighted) layer is forced to
     fallback
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -266,10 +270,13 @@ def forward(self, x):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18(ir):
+def test_resnet18(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     model = models.resnet18().eval().cuda()
     input = torch.randn((1, 3, 224, 224)).to("cuda")
 
@@ -310,12 +317,14 @@ def test_resnet18(ir):
 
 
 @pytest.mark.unit
-def test_hybrid_conv_fallback(ir):
+def test_hybrid_conv_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a hybrid
     model where a conv (a weighted layer)  has been forced to fallback to Pytorch.
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -372,13 +381,15 @@ def forward(self, x):
 
 
 @pytest.mark.unit
-def test_arange_export(ir):
+def test_arange_export(ir, tmpdir):
     """
     This tests export save and load functionality on a arange static graph
     Here the arange output is a static constant (which is registered as input to the graph)
     in the exporter.
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class MyModule(torch.nn.Module):
         def __init__(self):
             super().__init__()
@@ -436,10 +447,12 @@ def forward(self, x):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18_dynamic(ir):
+def test_resnet18_dynamic(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model with dynamic shapes
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
     model = models.resnet18().eval().cuda()
     input_bs2 = torch.randn((2, 3, 224, 224)).to("cuda")
 
@@ -510,10 +523,12 @@ def test_resnet18_dynamic(ir):
     not importlib.util.find_spec("torchvision"),
     "torchvision is not installed",
 )
-def test_resnet18_dynamic_fallback(ir):
+def test_resnet18_dynamic_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on Resnet18 model with dynamic shapes and fallback
     """
+
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
     model = models.resnet18().eval().cuda()
     input_bs2 = torch.randn((2, 3, 224, 224)).to("cuda")
 
@@ -584,11 +599,13 @@ def test_resnet18_dynamic_fallback(ir):
 
 
 @pytest.mark.unit
-def test_bitwise_and_dynamic_fallback(ir):
+def test_bitwise_and_dynamic_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a bitwise_and model with dynamic shapes and fallback
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class bitwise_and(torch.nn.Module):
         def forward(self, lhs_val, rhs_val):
             return torch.ops.aten.bitwise_and.Tensor(lhs_val, rhs_val)
@@ -664,11 +681,13 @@ def forward(self, lhs_val, rhs_val):
 
 
 @pytest.mark.unit
-def test_random_dynamic_fallback(ir):
+def test_random_dynamic_fallback(ir, tmpdir):
     """
     This tests export save and load functionality on a random model with dynamic shapes and fallback
     """
 
+    trt_ep_path = os.path.join(tmpdir, "trt.ep")
+
     class NeuralNetwork(nn.Module):
         def __init__(self):
             super(NeuralNetwork, self).__init__()
diff --git a/tests/py/dynamo/runtime/test_004_weight_streaming.py b/tests/py/dynamo/runtime/test_004_weight_streaming.py
index ac6b43730d..fe0ae649bc 100644
--- a/tests/py/dynamo/runtime/test_004_weight_streaming.py
+++ b/tests/py/dynamo/runtime/test_004_weight_streaming.py
@@ -6,7 +6,6 @@
 import torch_tensorrt as torchtrt
 from parameterized import parameterized
 from torch.testing._internal.common_utils import TestCase, run_tests
-from torch_tensorrt._utils import is_tegra_platform
 from torch_tensorrt.dynamo.utils import prepare_inputs
 
 INPUT_SIZE = (64, 100)
@@ -32,10 +31,6 @@ def forward(self, x):
         return out
 
 
-@unittest.skipIf(
-    is_tegra_platform(),
-    "Skipped on Tegra platforms",
-)
 class TestWeightStreamingPython(TestCase):
     @parameterized.expand(
         [
diff --git a/tests/py/dynamo/runtime/test_output_allocator.py b/tests/py/dynamo/runtime/test_output_allocator.py
index ce00135c99..c94020705c 100644
--- a/tests/py/dynamo/runtime/test_output_allocator.py
+++ b/tests/py/dynamo/runtime/test_output_allocator.py
@@ -5,7 +5,6 @@
 import torch_tensorrt
 from parameterized import parameterized
 from torch.testing._internal.common_utils import TestCase, run_tests
-from torch_tensorrt._utils import is_tegra_platform, is_thor
 
 from ..testing_utilities import DECIMALS_OF_AGREEMENT
 
@@ -45,7 +44,7 @@ def forward(self, input):
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "TensorRT RTX does not support nonzero which are required for this test",
 )
 class TestOutputAllocatorStaticModel(TestCase):
@@ -158,7 +157,7 @@ def test_combination_of_cg_and_oa(self, _, use_python_runtime):
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "TensorRT RTX does not support nonzero which are required for this test",
 )
 class TestOutputAllocatorDDSModel(TestCase):
@@ -268,7 +267,7 @@ def test_combination_of_cg_and_oa(self, _, use_python_runtime):
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "TensorRT RTX does not support nonzero which are required for this test",
 )
 class TestOutputAllocatorDDSOpWithReductionOpModel(TestCase):
@@ -382,7 +381,7 @@ def test_combination_of_cg_and_oa(self, _, use_python_runtime):
 
 
 @unittest.skipIf(
-    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx or is_thor() or is_tegra_platform(),
+    torch_tensorrt.ENABLED_FEATURES.tensorrt_rtx,
     "TensorRT RTX does not support nonzero which are required for this test",
 )
 class TestOutputAllocatorDDSModelWithGraphBreak(TestCase):