Fix TRT UT failures

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_elementwise.py

@@ -382,7 +382,7 @@ def sample_program_configs(self):
         def generate_input(shape):
             return np.random.random(shape).astype(np.float32)
 
-        for shape in [[4], [4, 32], [2, 64, 32], [1, 8, 16, 32]]:
+        for shape in [[4], [4, 32], [2, 32, 16], [1, 8, 16, 32]]:
             for op_type in [
                 "elementwise_add",
                 "elementwise_mul",
@@ -464,8 +464,8 @@ def generate_dynamic_shape(attrs):
                     "input_data2": [128, 128, 256],
                 }
                 self.dynamic_shape.opt_input_shape = {
-                    "input_data1": [2, 64, 64],
-                    "input_data2": [2, 64, 64],
+                    "input_data1": [2, 32, 16],
+                    "input_data2": [2, 32, 16],
                 }
             elif self.dims == 4:
                 self.dynamic_shape.min_input_shape = {

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_group_norm.py

@@ -129,7 +129,7 @@ def generate_trt_nodes_num(attrs, dynamic_shape):
         self.trt_param.precision = paddle_infer.PrecisionType.Half
         yield self.create_inference_config(), generate_trt_nodes_num(
             attrs, False
-        ), 1e-3
+        ), (1e-3, 1e-3)
 
         # for dynamic_shape
         generate_dynamic_shape(attrs)
@@ -140,7 +140,7 @@ def generate_trt_nodes_num(attrs, dynamic_shape):
         self.trt_param.precision = paddle_infer.PrecisionType.Half
         yield self.create_inference_config(), generate_trt_nodes_num(
             attrs, True
-        ), 1e-3
+        ), (1e-3, 1e-3)
 
     def add_skip_trt_case(self):
         pass

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_pool2d.py

@@ -20,6 +20,7 @@
 from typing import Any, Dict, List
 import unittest
 import itertools
+import copy
 
 
 class TrtConvertPool2dTest(TrtLayerAutoScanTest):
@@ -188,6 +189,39 @@ def teller(program_config, predictor_config):
             "The results of some cases are Nan, but the results of TensorRT and GPU are the same.",
         )
 
+    def assert_tensors_near(
+        self,
+        atol: float,
+        rtol: float,
+        tensor: Dict[str, np.array],
+        baseline: Dict[str, np.array],
+    ):
+        for key, arr in tensor.items():
+            self.assertEqual(
+                baseline[key].shape,
+                arr.shape,
+                'The output shapes are not equal, the baseline shape is '
+                + str(baseline[key].shape)
+                + ', but got '
+                + str(arr.shape),
+            )
+
+            # The result of Pool2d may have some elements that is the least value (-65504 for FP16),
+            # but for FP32 and FP16 precision, their least value are different.
+            # We set a threshold that is the least value of FP16,
+            # and make the values less than the threshold to be the threshold.
+            def align_less_threshold(arr, threshold):
+                return np.clip(arr, threshold, None)
+
+            fp16_min = np.finfo(np.float16).min
+            baseline_threshold = align_less_threshold(
+                copy.deepcopy(baseline[key]), fp16_min
+            )
+            arr_threshold = align_less_threshold(copy.deepcopy(arr), fp16_min)
+            np.testing.assert_allclose(
+                baseline_threshold, arr_threshold, rtol=rtol, atol=atol
+            )
+
     def test(self):
         self.add_skip_trt_case()
         self.run_test()