推理示例代码的 deploy/python/infer.py 推理 tood\\tood_r50_fpn_1x_coco模型大部分coco2007图片报错，少部分可以正常

问题确认 Search before asking

[X] 我已经查询历史issue，没有发现相似的bug。I have searched the issues and found no similar bug report.

Bug组件 Bug Component

Inference, Deploy

Bug描述 Describe the Bug

"tood\tood_r50_fpn_1x_coco" 推理使用的导出tood\tood_r50_fpn_1x_coco的模型为推理模型 tood\tood_r50_fpn_1x_coco 推理配置文件为 infer_cfg.yml, 内容如下, labellist根据实际整理出来了，未贴出 mode: paddle draw_threshold: 0.5 metric: COCO use_dynamic_shape: false arch: TOOD min_subgraph_size: 5 Preprocess:

keep_ratio: true target_size:
- 640
- 640 type: Resize
is_scale: true mean:
- 0.485
- 0.456
- 0.406 std:
- 0.229
- 0.224
- 0.225 type: NormalizeImage
type: Permute
stride: 32 type: PadStride label_list: []

""" File "E:\work_aiplat\pp_detection_sample\PaddleDetection\deploy\python\infer.py", result = self.predict() File "E:\work_aiplat\pp_detection_sample\PaddleDetection\deploy\python\infer.py", line xxx, in predict self.predictor.run() ValueError: In user code:

File "cus_share/PaddleDetection/eyecool_export_test_for_folder.py", line 155, in <module>
  export_pretrain_model_by_config_folder(path=WEIGHTS_HOME)
File "cus_share/PaddleDetection/eyecool_export_test_for_folder.py", line 40, in export_pretrain_model_by_config_folder
  export_pretrain_model_by_config_folder(item_path, suffix, outFilter)
File "cus_share/PaddleDetection/eyecool_export_test_for_folder.py", line 52, in export_pretrain_model_by_config_folder
  export_model_by_config_and_weight(item_path)
File "cus_share/PaddleDetection/eyecool_export_test_for_folder.py", line 148, in export_model_by_config_and_weight
  trainer.export(infer_folder)
File "/xxx/cus_share/PaddleDetection/ppdet/engine/trainer.py", line 734, in export
  save_dir)
File "/xxx/cus_share/PaddleDetection/ppdet/engine/trainer.py", line 711, in _get_infer_cfg_and_input_spec
  input_spec, static_model.forward.main_program,
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 563, in main_program
  concrete_program = self.concrete_program
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 479, in concrete_program
  return self.concrete_program_specify_input_spec(input_spec=None)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 517, in concrete_program_specify_input_spec
  *desired_input_spec)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 427, in get_concrete_program
  concrete_program, partial_program_layer = self._program_cache[cache_key]
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 723, in __getitem__
  self._caches[item] = self._build_once(item)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 714, in _build_once
  **cache_key.kwargs)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/decorator.py", line 232, in fun
  return caller(func, *(extras + args), **kw)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
  return wrapped_func(*args, **kwargs)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/base.py", line 51, in __impl__
  return func(*args, **kwargs)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 662, in from_func_spec
  outputs = static_func(*inputs)
File "/tmp/tmpp2e7y5ht.py", line 100, in forward
  false_fn_5, (), (inputs, self), (out,))
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 211, in convert_ifelse
  out = _run_py_ifelse(pred, true_fn, false_fn, true_args, false_args)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 257, in _run_py_ifelse
  return true_fn(*true_args) if pred else false_fn(*false_args)
File "/tmp/tmpp2e7y5ht.py", line 83, in false_fn_5
  for_loop_body_0, [__for_loop_var_index_0, inputs_list])
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 45, in convert_while_loop
  loop_vars = _run_py_while(cond, body, loop_vars)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/dygraph_to_static/convert_operators.py", line 59, in _run_py_while
  loop_vars = body(*loop_vars)
File "/tmp/tmpp2e7y5ht.py", line 78, in for_loop_body_0
  dy2static.convert_call(self.get_pred)())
File "/xxx/cus_share/PaddleDetection/ppdet/modeling/architectures/tood.py", line 75, in get_pred
  bbox_pred, bbox_num = self._forward()
File "/xxx/cus_share/PaddleDetection/ppdet/modeling/architectures/tood.py", line 62, in _forward
  head_outs = self.head(fpn_feats)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 917, in __call__
  return self._dygraph_call_func(*inputs, **kwargs)
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 907, in _dygraph_call_func
  outputs = self.forward(*inputs, **kwargs)
File "/xxx/cus_share/PaddleDetection/ppdet/modeling/heads/tood_head.py", line 255, in forward
  reg_dist = scale_reg(self.tood_reg(reg_feat).exp())
File "/xxx/cus_share/PaddleDetection/ppdet/modeling/bbox_utils.py", line 766, in batch_distance2bbox
  # while tensor add parameters, parameters should be better placed on the second place
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/layers/math_op_patch.py", line 347, in __impl__
  attrs={'axis': axis})
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/framework.py", line 3184, in append_op
  attrs=kwargs.get("attrs", None))
File "/xxx/condas/envs/xxx/lib/python3.7/site-packages/paddle/fluid/framework.py", line 2224, in __init__
  for frame in traceback.extract_stack():

InvalidArgumentError: Broadcast dimension mismatch. Operands could not be broadcast together with the shape of X = [1, 4480, 2] and the shape of Y = [1, 6400, 2]. Received [4480] in X is not equal to [6400] in Y at i:1.
  [Hint: Expected x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1 == true, but received x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1:0 != true:1.] (at C:\home\workspace\Paddle_release\paddle/fluid/operators/elementwise/elementwise_op_function.h:240)
  [operator < elementwise_add > error]
  """

复现环境 Environment

OS: Linux/Windows PaddlePaddle: 2.2.2 PaddleDetection: release/2.5 Python: 3.7.0 CUDA: 10.1 CUDNN: 7.6 GCC: 8.2.0 windows使用的cpu推理的，linux使用cpu/gpu 均出现该错误

Bug描述确认 Bug description confirmation

[X] 我确认已经提供了Bug复现步骤、代码改动说明、以及环境信息，确认问题是可以复现的。I confirm that the bug replication steps, code change instructions, and environment information have been provided, and the problem can be reproduced.

是否愿意提交PR？ Are you willing to submit a PR?

[X] 我愿意提交PR！I'd like to help by submitting a PR!

PaddlePaddle / PaddleDetection