I'm using NCS2 on Windows 10 64 bit, my model is designed by tensorflow. The model that I'm using is RL, when converting to IR from .PB file, I selected output is "import/primaryQN/norm/Squeeze" and converted sucessfully.
This's my script :
python mo_tf.py -m D3RQN_phase3.pb --data_type FP16 --input "import/primaryQN/preContextsFlattened[1 1 248],import/primaryQN/LSTMCellZeroState/zeros[1 248],import/primaryQN/LSTMCellZeroState/zeros_1[1 248]" --output import/primaryQN/norm/Squeeze,import/primaryQN/LSTM_hidden_cell_output,import/primaryQN/LSTM_hidden_state_output
(You can follow the image)
I run IR on VPU is cannot but it's work when run on CPU.
My question is
Is layer name Squeeze not supported run on VPU?
If not, Is it any difference layer with the same but can run on VPU?
Thanks for reaching out.
The Squeeze layer is supported by the MYRIAD plugin, as stated in the Supported Layers documentation.
Could you please give us the following for us to test this on our end:
- Frozen model
- IR files
- Sample / demo to run inference
- Expected output