Hi There,
I'm currently trying to convert a Tensorflow model to a Neural Compute Stick which contains a multi-dimensional reduction block, using mo.py
I experience the following error:
[ ERROR ] -------------------------------------------------
[ ERROR ] ----------------- INTERNAL ERROR ----------------
[ ERROR ] Unexpected exception happened.
[ ERROR ] Please contact Model Optimizer developers and forward the following information:
[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID (<class 'extensions.middle.ConvertGroupedStridedSlice.ConvertGroupedStridedSlice'>)": After re-inference of conv2_block1_2_reduce/strided_slice/Split node, old and new shapes do not match. Old shapes: [array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1])], new shapes: [array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4])].
[ ERROR ] Traceback (most recent call last):
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/utils/class_registration.py", line 273, in apply_replacements
for_graph_and_each_sub_graph_recursively(graph, replacer.find_and_replace_pattern)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/middle/pattern_match.py", line 58, in for_graph_and_each_sub_graph_recursively
func(graph)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/extensions/middle/ConvertGroupedStridedSlice.py", line 205, in find_and_replace_pattern
split.create_node_with_data(inputs=[input_data], data_nodes=final_data_nodes_list)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/ops/op.py", line 257, in create_node_with_data
[data_node.shape for data_node in data_nodes])
AssertionError: After re-inference of conv2_block1_2_reduce/strided_slice/Split node, old and new shapes do not match. Old shapes: [array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1])], new shapes: [array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4])].
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/main.py", line 302, in main
return driver(argv)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/main.py", line 251, in driver
is_binary=not argv.input_model_is_text)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/pipeline/tf.py", line 134, in tf2nx
class_registration.apply_replacements(graph, class_registration.ClassType.MIDDLE_REPLACER)
File "/opt/intel/openvino_2019.2.242/deployment_tools/model_optimizer/mo/utils/class_registration.py", line 299, in apply_replacements
)) from err
Exception: Exception occurred during running replacer "REPLACEMENT_ID (<class 'extensions.middle.ConvertGroupedStridedSlice.ConvertGroupedStridedSlice'>)": After re-inference of conv2_block1_2_reduce/strided_slice/Split node, old and new shapes do not match. Old shapes: [array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1]), array([ 1, 56, 56, 32, 1])], new shapes: [array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4]), array([ 1, 56, 56, 32, 1, 4])].
[ ERROR ] ---------------- END OF BUG REPORT --------------
[ ERROR ] -------------------------------------------------
Having had a look through the stack trace (split.create_node_with_data(inputs=[input_data], data_nodes=final_data_nodes_list) from ConvertGroupedStridedSlice.py (approx. line 200) -> create_node_with_data from model_optimizer/mo/ops/op.py (line 157) -> to line 185 - it seems I've hit the 'TODO' case of having multiple output ports. If we see the model summery and relevant part of the graph I'm using:
...
conv2_block1_2_conv (DepthwiseC (None, 56, 56, 512) 4608 conv2_block1_2_pad[0][0] __________________________________________________________________________________________________
reshape_1 (Reshape) (None, 56, 56, 32, 4 0 conv2_block1_2_conv[0][0] __________________________________________________________________________________________________ conv2_block1_2_reduce (Lambda) (None, 56, 56, 32, 4 0 reshape_1[0][0] __________________________________________________________________________________________________
reshape_2 (Reshape) (None, 56, 56, 128) 0 conv2_block1_2_reduce[0][0] __________________________________________________________________________________________________
...
For greater clarity, here are the relevant sub-graphs from tensorboard:
Since "old_data_shape" is [ 1 56 56 32 1], rather than the expected [ 1 56 56 32 4], there is the assertion error. Is this due to the editing/removing of connections from the old_data_nodes (in ConvertGroupedStridedSlice.py from lines 182)?
Any help would be greatly appreciated!
Link Copied
