paddlehub ultra_light_fast_generic_face_detector_1mb_640转onnx报错

[w-Bro]

请将下面信息填写完整，便于我们快速解决问题，谢谢！

模型路径为paddlehub目录下的ultra_light_fast_generic_face_detector_1mb_640 ，是否需要进行其他转换？

命令：paddle2onnx --model_dir ./ultra_light_fast_generic_face_detector_1mb_640 --save_file onnx_file --opset_version 10 --enable_onnx_checker True
版本：paddle2onnx 0.9.0、onnx 1.9.0
问题描述
请在此处详细的描述报错信息

Exception: Error happened when mapping node ['conv2d_0'] to onnx, which op_type is 'conv2d' with inputs: {'Bias': [], 'Filter': ['_base_net_0_0_weight'], 'Input': ['_input'], 'ResidualData': []} and outputs: {'Output': ['_245.tmp_0']}, specific error: The conv data format should be 'NCHW' or 'NCDHW', but received data format is AnyLayout.

更多信息 :

• 用于部署的推理引擎:
• 为什么需要转换为ONNX格式：
• Paddle2ONNX版本:
• 你的联系方式(Email/Wechat/Phone):

报错截图

其他信息

[jiangjiajun]

@w-Bro Hi，请安装github上最新代码重新再导出onnx试下

pip uninstall paddle2onnx
git clone https://github.com/PaddlePaddle/Paddle2ONNX.git
cd Paddle2ONNX
python setup.py install

另外我们也准备了一份问卷，方便的话请抽出半分钟时间帮完成问卷 https://iwenjuan.baidu.com/?code=r8hu2s 帮助Paddle2ONNX做得更好

[w-Bro]

还是不行，报了新的错误。

Traceback (most recent call last):
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\op_mapper\op_mapper.py", line 119, in mapping
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\op_mapper\tensor.py", line 35, in opset_4
TypeError: object of type 'NoneType' has no len()

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "D:\Miniconda3\envs\paddle212-py37\Scripts\paddle2onnx-script.py", line 33, in <module>
    sys.exit(load_entry_point('paddle2onnx==0.9.0', 'console_scripts', 'paddle2onnx')())
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\command.py", line 195, in main
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\command.py", line 159, in program2onnx
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\convert.py", line 88, in program2onnx
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\convert.py", line 36, in export_onnx
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\graph\onnx_graph.py", line 262, in build
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\graph\onnx_graph.py", line 209, in build_op_nodes
  File "D:\Miniconda3\envs\paddle212-py37\lib\site-packages\paddle2onnx-0.9.0-py3.7.egg\paddle2onnx\op_mapper\op_mapper.py", line 124, in mapping
Exception: Error happened when mapping node ['concat_0'] to onnx, which op_type is 'concat' with inputs: {'X': ['_290.tmp_2', '_297.tmp_2', '_307.tmp_2']} and outputs: {'Out': ['concat_0.tmp_0']}, specific error: object of type 'NoneType' has no len()

[neonhuang]

你好，这个问题我修复了，你可以使用develop分支
PR：#510
这是因为concat中，node.input('AxisTensor')可能是None;

你可以将paddle2onnx/op_mapper/tensor.py的35行修改如下一下：
node_axis = node.input('AxisTensor')
if node_axis is not None and len(node_axis) > 0:
。。。。

[w-Bro]

可以了，谢谢