MaxPool3D not supported

[ralovich]

Describe the bug
MaxPool3D layers not supported

Urgency
If there are particular important use cases blocked by this or strict project-related timelines, please share more information and dates. If there are no hard deadlines, please specify none.

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubunut 20.04
• Tensorflow Version: 2.2
• Python version: 3.8

To Reproduce
Reproducer network available at https://github.com/ralovich/trt-regression
$python -m tf2onnx.convert --input model_100.pb --inputs input_1_100_100_100_1:0 --outputs inception_v3_3d_1/output/sub_1:0 --output model_100.onnx --opset 12

Expected behavior
Also 3d pooling (AvgPool3D and MaxPool3D) layers are converted.

Screenshots
INFO:tensorflow:Froze 0 variables.
2020-07-23 18:47:46,076 - INFO - Froze 0 variables.
INFO:tensorflow:Converted 0 variables to const ops.
2020-07-23 18:47:46,080 - INFO - Converted 0 variables to const ops.
2020-07-23 18:47:46,554 - INFO - Using tensorflow=2.2.0, onnx=1.7.0, tf2onnx=1.7.0/38b1a6
2020-07-23 18:47:46,554 - INFO - Using opset <onnx, 12>
2020-07-23 18:47:47,219 - ERROR - Tensorflow op [inception_v3_3d_1/full_conv/pool2: MaxPool3D] is not supported
2020-07-23 18:47:47,252 - ERROR - Unsupported ops: Counter({'MaxPool3D': 1})
2020-07-23 18:47:47,257 - INFO - Optimizing ONNX model
2020-07-23 18:47:47,334 - WARNING - Failed to apply remove_back_to_back
Traceback (most recent call last):
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/init.py", line 52, in optimize_graph
graph = opt.optimize(current) or graph
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/optimizer_base.py", line 41, in optimize
graph = self._optimize(graph)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 35, in _optimize
return self._apply_optimization(graph, self._optimize_at_current_graph_level)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/optimizer_base.py", line 62, in _apply_optimization
graph = optimize_func(graph)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 67, in _optimize_at_current_graph_level
q2 = handler(g, node, consumer_nodes)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 134, in _optimize_transpose
new_perm = [t1[i] for i in t2]
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 134, in
new_perm = [t1[i] for i in t2]
IndexError: list index out of range
2020-07-23 18:47:47,391 - WARNING - Failed to apply remove_back_to_back
Traceback (most recent call last):
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/init.py", line 52, in optimize_graph
graph = opt.optimize(current) or graph
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/optimizer_base.py", line 41, in optimize
graph = self._optimize(graph)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 35, in _optimize
return self._apply_optimization(graph, self._optimize_at_current_graph_level)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/optimizer_base.py", line 62, in _apply_optimization
graph = optimize_func(graph)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 67, in _optimize_at_current_graph_level
q2 = handler(g, node, consumer_nodes)
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 134, in _optimize_transpose
new_perm = [t1[i] for i in t2]
File "/home/tade/venv-tf-2.2/lib/python3.8/site-packages/tf2onnx/optimizer/back_to_back_optimizer.py", line 134, in
new_perm = [t1[i] for i in t2]
IndexError: list index out of range
2020-07-23 18:47:47,393 - INFO - After optimization: Const -22 (63->41), Identity -1 (1->0), Transpose -4 (38->34)
2020-07-23 18:47:47,398 - INFO -
2020-07-23 18:47:47,398 - INFO - Successfully converted TensorFlow model model_100.pb to ONNX
2020-07-23 18:47:47,404 - INFO - ONNX model is saved at model_100.onnx

Additional context
Add any other context about the problem here. If the issue is about a particular model, please share the model details as well to facilitate debugging.

[ralovich]

Maybe the network in the above reproducer (https://github.com/ralovich/trt-regression) could be used to build a test-case for this?

[TomWildenhain-Microsoft]

Working on it with #1020