MXNet official tutorial model to tensorflow

[ShoufaChen]

Platform (like ubuntu 16.04/win10): ubuntu16.04 Anaconda

Python version:3.6

Source framework with version (like Tensorflow 1.4.1 with GPU):1.8 with GPU

Destination framework with version (like CNTK 2.3 with GPU): MXNet 1.2.1 with GPU

Pre-trained model path (webpath or webdisk path): https://drive.google.com/drive/folders/1MQUYEPYnvWRmGBPh338Le5DFYnq-yr7Z?usp=sharing

(this model is got from the MXNet official tutorial Saving and Loading Gluon Models
Running scripts:

 python -m mmdnn.conversion._script.convertToIR -f mxnet -n lenet-symbol.json -w lenet-0001.params -d resnet --inputShape 1 28 28

Error:

el_names=['softmax_label']) but input with name 'softmax_label' is not found in symbol.list_arguments(). Did you mean one of:
        data
  warnings.warn(msg)
Warning: MXNet Parser has not supported operator null with name data.
Warning: convert the null operator with name [data] into input layer.
infer_shape error. Arguments:
  data: (1, 1)
Traceback (most recent call last):
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 199, in <module>
    _main()
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 194, in _main
    ret = _convert(args)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 117, in _convert
    parser.run(args.dstPath)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/common/DataStructure/parser.py", line 22, in run
    self.gen_IR()
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/mxnet/mxnet_parser.py", line 262, in gen_IR
    func(current_node)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/mxnet/mxnet_parser.py", line 424, in rename_Convolution
    self.set_output_shape(source_node, IR_node)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/mxnet/mxnet_parser.py", line 281, in set_output_shape
    arg_shape, output_shape, aux_shape = sym.infer_shape(data = self.data_shape)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/symbol/symbol.py", line 990, in infer_shape
    res = self._infer_shape_impl(False, *args, **kwargs)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/symbol/symbol.py", line 1120, in _infer_shape_impl
    ctypes.byref(complete)))
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/base.py", line 149, in check_call
    raise MXNetError(py_str(_LIB.MXGetLastError()))
mxnet.base.MXNetError: Error in operator hybridsequential0_conv0_fwd: [09:51:51] src/operator/nn/convolution.cc:143: Check failed: dshp.ndim() == 4U (2 vs. 4) Input data should be 4D in batch-num_filter-y-x

Stack trace returned 10 entries:
[bt] (0) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(+0x31f81a) [0x7f03434c581a]
[bt] (1) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(+0x31fe41) [0x7f03434c5e41]
[bt] (2) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(+0x54a1e1) [0x7f03436f01e1]
[bt] (3) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(+0x24de24a) [0x7f034568424a]
[bt] (4) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(+0x24e0b40) [0x7f0345686b40]
[bt] (5) /home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/libmxnet.so(MXSymbolInferShape+0x15aa) [0x7f03456008ba]
[bt] (6) /home/csf/anaconda3/envs/tf18/lib/python3.6/lib-dynload/../../libffi.so.6(ffi_call_unix64+0x4c) [0x7f0371dc1ec0]
[bt] (7) /home/csf/anaconda3/envs/tf18/lib/python3.6/lib-dynload/../../libffi.so.6(ffi_call+0x22d) [0x7f0371dc187d]
[bt] (8) /home/csf/anaconda3/envs/tf18/lib/python3.6/lib-dynload/_ctypes.cpython-36m-x86_64-linux-gnu.so(_ctypes_callproc+0x2ce) [0x7f0371fd6e2e]
[bt] (9) /home/csf/anaconda3/envs/tf18/lib/python3.6/lib-dynload/_ctypes.cpython-36m-x86_64-linux-gnu.so(+0x12865) [0x7f0371fd7865]

[namizzz]

Hi @ShoufaChen ,The error shows "Input data should be 4D in batch-num_filter-y-x". you may need to use:

--inputShape 1,28,28
not
--inputShape 1 28 28

[ShoufaChen]

After change the `--inputShaep`` I met another queation:

Running scripts:

python -m mmdnn.conversion._script.convertToIR -f mxnet -n lenet-symbol.json -w lenet-0001.params -d tflenet --inputShape 1,28,28

Error:

/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mxnet/module/base_module.py:54: UserWarning: You created Module with Module(..., label_names=['softmax_label']) but input with name 'softmax_label' is not found in symbol.list_arguments(). Did you mean one of:
        data
  warnings.warn(msg)
Warning: MXNet Parser has not supported operator null with name data.
Warning: convert the null operator with name [data] into input layer.
Traceback (most recent call last):
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 199, in <module>
    _main()
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 194, in _main
    ret = _convert(args)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 117, in _convert
    parser.run(args.dstPath)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/common/DataStructure/parser.py", line 22, in run
    self.gen_IR()
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/mxnet/mxnet_parser.py", line 262, in gen_IR
    func(current_node)
  File "/home/csf/anaconda3/envs/tf18/lib/python3.6/site-packages/mmdnn/conversion/mxnet/mxnet_parser.py", line 495, in rename_Convolution
    weight = self.weight_data.get(source_node.name + "_weight").asnumpy()
AttributeError: 'NoneType' object has no attribute 'asnumpy

[namizzz]

Hi @ShoufaChen , if you are using Gluon not the original MXNet. The name of weights and bias is different . You can refer to this #208.

[hellolingling]

@ShoufaChen I also have the same problem as you. When changing the -- input shape as 3,112,112, I also got the same problem as you. Have you fixed your problem?

[namizzz]

Hello @hellolingling , if you are using Gluon not the original MXNet. The name of weights and bias is different . You can refer to this #208

[hellolingling]

Hi, @namizzz thank you for your reply!
But, I referred to #208, and changed the code as suggested. And I am not using Gluon. I met the same problem as shown below:
Input：
python -m mmdnn.conversion._script.convertToIR -f mxnet -n model-symbol.json -w model-0000.params -d model-descr512 --inputShape 3,112,112

<mmdnn.conversion.mxnet.mxnet_parser.MXNetParser object at 0x7fdd6b129f10> Warning: MXNet Parser has not supported operator null with name data. Warning: convert the null operator with name [data] into input layer. Traceback (most recent call last): File "/home/zhulingling/anaconda2/lib/python2.7/runpy.py", line 174, in _run_module_as_main "__main__", fname, loader, pkg_name) File "/home/zhulingling/anaconda2/lib/python2.7/runpy.py", line 72, in _run_code exec code in run_globals File "/home/zhulingling/workspace/model_conversion/MMdnn/mmdnn/conversion/_script/convertToIR.py", line 207, in <module> _main() File "/home/zhulingling/workspace/model_conversion/MMdnn/mmdnn/conversion/_script/convertToIR.py", line 201, in _main ret = _convert(args) File "/home/zhulingling/workspace/model_conversion/MMdnn/mmdnn/conversion/_script/convertToIR.py", line 121, in _convert parser.run(args.dstPath) File "mmdnn/conversion/common/DataStructure/parser.py", line 22, in run self.gen_IR() File "mmdnn/conversion/mxnet/mxnet_parser.py", line 264, in gen_IR func(current_node) File "mmdnn/conversion/mxnet/mxnet_parser.py", line 400, in rename_FullyConnected weight = self.weight_data.get(source_node.name.replace("fwd","weight")).asnumpy().transpose((1, 0)) AttributeError: 'NoneType' object has no attribute 'asnumpy'
Can you help solve the problem?

[hellolingling]

Hi, @namizzz ,according the issue https://github.com/Microsoft/MMdnn/issues/231.
I changed the
File "mmdnn/conversion/mxnet/mxnet_parser.py", line 401, in rename_FullyConnected weight = self.weight_data.get(source_node.name + "_weight").asnumpy().transpose((1, 0))
into
weight = self.weight_data.get("fc1_weight").asnumpy().transpose((1, 0)).
This worked, thank you!

[hellolingling]

refer to #231