OpenVINO GNMT

[IDZ_A_Intel]

Hi,

I am trying to get GNMT running on a CPU with OpenVINO.

I followed the instructions in the below link exactly:

https://docs.openvinotoolkit.org/2019_R3.1/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_GNMT_From_Tensorflow.html#patch-file

However, when I run the following command:

python3 benchmark_app.py -i test.txt -m frozen_GNMT_inference_graph.xml -d CPU

I see this error:

[ ERROR ] Unsupported primitive of type: Unsqueeze name: dynamic_seq2seq/decoder/tile_batch_1/ExpandDims
Traceback (most recent call last):
  File "/opt/intel/openvino_2019.3.376/deployment_tools/tools/benchmark_tool/benchmark_app.py", line 82, in main
    exe_network = benchmark.load_network(ie_network, perf_counts, args.number_infer_requests)
  File "/opt/intel/openvino_2019.3.376/python/python3.6/openvino/tools/benchmark/benchmark.py", line 127, in load_network
    num_requests=number_infer_requests or 0)
  File "ie_api.pyx", line 85, in openvino.inference_engine.ie_api.IECore.load_network
  File "ie_api.pyx", line 92, in openvino.inference_engine.ie_api.IECore.load_network
RuntimeError: Unsupported primitive of type: Unsqueeze name: dynamic_seq2seq/decoder/tile_batch_1/ExpandDims
 

Please could I get help with this.

Also, when I want to run GNMT using the Python API, is there a sample as to how to generate the 'input_data_tensor' and 'seq_lengths' fields for the following piece of code:

# Creating input data

input_data = {'IteratorGetNext/placeholder_out_port_0': input_data_tensor,

'IteratorGetNext/placeholder_out_port_1': seq_lengths}

[Cary_P_Intel1]

Hi, Admin,

I tried the GNMT model from the github and reduplicate your error, I am checking with the develop team to know where the issue is, will update you later.

[Anonymous102]

Thanks for taking a look into this issue. Looking forward to your findings.

[Anonymous102]

Hi,

Please could you give out an update on the status of this issue.