09-10-2019 05:17 AM - edited 09-10-2019 05:23 AM
I am trying to compile Tensorflow's Conv2DBackpropInput using DNNC.
dnnc version v3.00
DPU Target : v1.4.0
Build Label: Aug 9 2019 13:19:12
I get the following error:
[DNNC][Error] 'Const' op should be fused with current op [Conv2DBackpropInput] by DECENT.
Please tell me how to fix it.
See below for compiled models.
The attached script "create_graph.py" is a file that generates a model.
The attached script "test_input_fn.py" is a decent_q calibration input.
This model's deconvolution is "tf.keras.layers.Conv2DTranspose".