How to convert F.interpolate to onnx?

deployment
1

I’ve converted an east model into ONNX，but when checking model by onnx.checker.check_model(), I got the following error:

Traceback (most recent call last):
  File "east_onnx_demo.py", line 223, in <module>
    onnx.checker.check_model(net)
  File "/home/dai/py36env/lib/python3.6/site-packages/onnx/checker.py", line 86, in check_model
    C.check_model(model.SerializeToString())
onnx.onnx_cpp2py_export.checker.ValidationError: Node () has input size 4 not in range [min=2, max=2].

==> Context: Bad node spec: input: "901" input: "924" input: "924" input: "923" output: "925" op_type: "Resize" attribute { name: "coordinate_transformation_mode" s: "pytorch_half_pixel" type: STRING } attribute { name: "cubic_coeff_a" f: -0.75 type: FLOAT } attribute { name: "mode" s: "linear" type: STRING } attribute { name: "nearest_mode" s: "floor" type: STRING }

The problem could come from the following code:

def forward(self, x1,x2,x3,x4):
        y = F.interpolate(
            x4, scale_factor=2.0, mode="bilinear", align_corners=False
        )
        y = torch.cat((y, x3), 1)
        y = self.relu1(self.bn1(self.conv1(y)))
        y = self.relu2(self.bn2(self.conv2(y)))

        y = F.interpolate(
            y, scale_factor=2.0, mode="bilinear", align_corners=False
        )
        y = torch.cat((y, x2), 1)
        y = self.relu3(self.bn3(self.conv3(y)))
        y = self.relu4(self.bn4(self.conv4(y)))

        y = F.interpolate(
            y, scale_factor=2.0, mode="bilinear", align_corners=False
        )
        y = torch.cat((y, x1), 1)
        y = self.relu5(self.bn5(self.conv5(y)))
        y = self.relu6(self.bn6(self.conv6(y)))

        y = self.relu7(self.bn7(self.conv7(y)))
        return y

Environment: Pytorch1.3.1, ONNX1.5.0

How can I solve this problem?

1 Like