How to compile QDQ model with onnxrt_ep.py

[wintersurvival]

Hi,

I'm attempting to compile an QDQ model with the config option:

        'optional_options' :
        {
            'advanced_options:prequantized_model': 1,
        }

but it seems the QDQ model is unsupported:

python onnxrt_ep.py -c

Available execution providers :  ['TIDLExecutionProvider', 'TIDLCompilationProvider', 'CPUExecutionProvider']
Running 1 Models - ['cl-dlr_mobilenetv3_large']
Running_Model :  cl-dlr_mobilenetv3_large  
Running shape inference on model mobilenetv3_large_qdq.onnx 

Preliminary subgraphs created = 30 

*** WARNING : Number of subgraphs generated > max_num_subgraphs provided in options - additional subgraphs are delegated to ARM *** 
Final number of subgraphs created are : 16, - Offloaded Nodes - 331, Total Nodes - 493 
Could not find const or initializer of layer /module/0_34/Conv !!!
Weight Tensor size is not matching with Proto kernel_shape 
Could not find const or initializer of layer /module/0_34/Conv !!!

Unsupported Onnx import data type : 0 
Could not find const or initializer of layer /module/0_34/Conv !!!

Unsupported Onnx import data type : 0 
Could not find const or initializer of layer /module/fc1_4/Conv !!!
Weight Tensor size is not matching with Proto kernel_shape 
Could not find const or initializer of layer /module/fc1_4/Conv !!!

Unsupported Onnx import data type : 0 
Could not find const or initializer of layer /module/fc1_4/Conv !!!

Unsupported Onnx import data type : 0 
Could not find const or initializer of layer /module/fc2_4/Conv !!!
Weight Tensor size is not matching with Proto kernel_shape 
Could not find const or initializer of layer /module/fc2_4/Conv !!!

How to compile the QDQ model?

@mathmanu @debapriyamaji @gibrane @kumardesappan @yuanzhaoti

Thank you!