Failed to convert Add/Mul/Pow op from onnx to acuity case code

i want to convert my onnx model to case code. i use the acuity-toolkit inside aml_npu_sdk_6.4.3. All demos run successfully. my env is numbqq/npu-sdk container, which has:

  • python 3.6.9
  • torch 1.2.0
  • tensorflow 1.13.2

i use the following code to generate a onnx model with one Add/Mul/Pow op.

import numpy as np
import torch

class TestNet(torch.nn.Module):
    def __init__(self):
        super(TestNet, self).__init__()
    def forward(self, x):
        # y = torch.pow(x,2)
        y = torch.add(x,x)
        # y = torch.mul(x,x)
        return y

if __name__ == '__main__':
    model = TestNet()
    dummy_input = torch.randn(1, 1, 360, 640)
    torch.onnx.export(model, dummy_input, "test.onnx")

then, i use the, but all failed with following output:

Any suggestions for this? Thanks.

any one can help me ?

same here. closed source converter really sucks