Fix several unit tests and increase the unit tests stability (#43670)

* Reduce gather op unit tests size and increase the timeout

* Add NVIDIA_TF32_OVERRIDE for multi-processes environment

* Remove record test for device event ut

paddle/fluid/platform/device_event_test.cc

@@ -44,8 +44,6 @@ TEST(DeviceEvent, CUDA) {
   ASSERT_EQ(status, true);
   // case 2. test for event_recorder
   event.Record(context);
-  status = event.Query();
-  ASSERT_EQ(status, false);
   // case 3. test for event_finisher
   event.Finish();
   status = event.Query();

python/paddle/fluid/tests/unittests/CMakeLists.txt

@@ -1356,7 +1356,7 @@ set_tests_properties(test_matmul_op PROPERTIES TIMEOUT 120)
 set_tests_properties(test_nearest_interp_v2_op PROPERTIES TIMEOUT 120)
 set_tests_properties(test_trilinear_interp_op PROPERTIES TIMEOUT 120)
 set_tests_properties(test_bicubic_interp_v2_op PROPERTIES TIMEOUT 120)
-set_tests_properties(test_gather_op PROPERTIES TIMEOUT 120)
+set_tests_properties(test_gather_op PROPERTIES TIMEOUT 180)
 set_tests_properties(test_static_save_load PROPERTIES TIMEOUT 250)
 set_tests_properties(test_pylayer_op PROPERTIES TIMEOUT 120)
 set_tests_properties(test_paddle_save_load_binary PROPERTIES TIMEOUT 120)

python/paddle/fluid/tests/unittests/test_collective_api_base.py

@@ -208,21 +208,15 @@ class TestDistBase(unittest.TestCase):
             with_gloo = '0'
         else:
             with_gloo = '1'
-        required_envs = {
-            "FLAGS_fraction_of_gpu_memory_to_use": "0.15",
-            "FLAGS_eager_delete_tensor_gb": "0.0",
-            "PATH": os.getenv("PATH"),
-            "PYTHONPATH": os.getenv("PYTHONPATH", ""),
-            "LD_LIBRARY_PATH": os.getenv("LD_LIBRARY_PATH", ""),
-            "LD_PRELOAD": os.getenv("LD_PRELOAD", ""),
-            "FLAGS_call_stack_level": "2",
-            "GLOG_v": "3",
+        required_envs = os.environ.copy()
+        additional_envs = {
             "NCCL_P2P_DISABLE": "1",
             "STATIC_MODE": static_mode,
             "PADDLE_WITH_GLOO": with_gloo,
             "BACKEND": backend,
             "PATH_ID": path_id
         }
+        required_envs.update(additional_envs)
         required_envs.update(need_envs)
         if check_error_log:
             required_envs["GLOG_v"] = "3"

python/paddle/fluid/tests/unittests/test_gather_op.py

@@ -317,7 +317,7 @@ class API_TestDygraphGather(unittest.TestCase):
             return
 
         x = np.random.rand(226862, 256).astype("float32")
-        index = np.random.randint(0, 22682, size=(11859027))
+        index = np.random.randint(0, 22682, size=(8859027))
 
         def test_dygraph():
             with fluid.dygraph.guard():