[https://nvbugs/5640873][fix] Move thop tests to pre-merge (NVIDIA#9094)

Signed-off-by: Hui Gao <huig@nvidia.com>

Author: HuiGao-NV

tests/integration/test_lists/test-db/l0_b300.yml

@@ -17,7 +17,8 @@ l0_b300:
   tests:
   # ------------- PyTorch tests ---------------
   - unittest/_torch/attention # 200s
-  # - unittest/_torch/thop
+  - unittest/_torch/thop/parallel TIMEOUT (90)
+  - unittest/_torch/thop/serial
   - unittest/_torch/executor # 250s
   - unittest/_torch/modules # 300s
   - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8B::test_nvfp4

tests/integration/test_lists/test-db/l0_dgx_b300.yml

@@ -16,7 +16,6 @@ l0_dgx_b300:
       backend: pytorch
   tests:
   - unittest/_torch/attention
-  - unittest/_torch/thop
   - unittest/_torch/executor
   - unittest/_torch/modules
   - unittest/_torch/modeling -k "modeling_llama"

tests/unittest/_torch/thop/parallel/test_mamba_conv1d_op.py

@@ -15,6 +15,9 @@ class TestFunctional(unittest.TestCase):
     def setUp(self):
         tensorrt_llm.logger.set_level('error')
 
+    def tearDown(self):
+        torch.cuda.empty_cache()
+
     @parameterized.expand(
         list(
             product([2048], [4], ['context', 'generation'],