KeyError Thrown trying to convert split Op

[zcin]

Platform: Ubuntu 16.04

Python version: 3.6.9

Source framework with version: Tensorflow 1.4.0 with CPU

Destination framework with version: PyTorch 1.1.0 with CPU

Pre-trained model path: N/A

Running scripts: mmtoir -f tensorflow -d vgg_output -n vgg.ckpt.meta -w vgg.ckpt --dstNodeName add_16

I am using a version of the vgg16 tensorflow model that has been converted from ONNX using onnx-tf. Within this model there are split operators, at which the converter throws an error:

Traceback (most recent call last):
  File "/home/mtc/anaconda3/envs/torch-onnx/bin/mmtoir", line 10, in <module>
    sys.exit(_main())
  File "/home/mtc/anaconda3/envs/torch-onnx/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 192, in _main
    ret = _convert(args)
  File "/home/mtc/anaconda3/envs/torch-onnx/lib/python3.6/site-packages/mmdnn/conversion/_script/convertToIR.py", line 115, in _convert
    parser.run(args.dstPath)
  File "/home/mtc/anaconda3/envs/torch-onnx/lib/python3.6/site-packages/mmdnn/conversion/common/DataStructure/parser.py", line 22, in run
    self.gen_IR()
  File "/home/mtc/anaconda3/envs/torch-onnx/lib/python3.6/site-packages/mmdnn/conversion/tensorflow/tensorflow_parser.py", line 424, in gen_IR
    func(current_node)
  File "/home/mtc/anaconda3/envs/torch-onnx/lib/python3.6/site-packages/mmdnn/conversion/tensorflow/tensorflow_parser.py", line 548, in rename_Conv2D
    self.set_weight(source_node.name, 'weights', self.ckpt_data[W.name])
KeyError: 'split/split_dim'

Does the converter currently support converting split operators?
Any help would be appreciated, thanks!

[JiahaoYao]

In the onnx model, the vgg network looks like this:

However, when converted to the tensorflow by onnx-tf, there are split operations.

However, I think split is not supported in MMdnn now.