如何转换为onnx模型 about elegant

[ZTMIDGO]

    device = 'cpu'
    c = torch.randn(1, 3, 256, 256, requires_grad=True).to(device)
    s = torch.randn(1, 3, 256, 256, requires_grad=True).to(device)

    mask_c = torch.randn(1, 2, 256, 256, requires_grad=True).to(device)
    mask_s = torch.randn(1, 2, 256, 256, requires_grad=True).to(device)

    diff_c = torch.randn(1, 136, 256, 256, requires_grad=True).to(device)
    diff_s = torch.randn(1, 136, 256, 256, requires_grad=True).to(device)

    lms_c = torch.randn(1, 68, 2, requires_grad=True).to(device)
    lms_s = torch.randn(1, 68, 2, requires_grad=True).to(device)

    input = ['c', 's', 'mask_c', 'mask_s', 'diff_c', 'diff_s', 'lms_c', 'lms_s']
    output = ['output']

    torch.onnx.export(self.solver.G, (c, s, mask_c, mask_s, diff_c, diff_s, lms_c, lms_s), 'model.onnx', verbose=True, 
    opset_version=15, input_names=input, output_names=output)

无法使用以上代码转换为onnx：
Exporting the operator inverse to ONNX opset version 15 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub.

希望可以指教一下如何解决问题。

[Fcucgvhhhvjv]

@ZTMIDGO

    device = 'cpu'
    c = torch.randn(1, 3, 256, 256, requires_grad=True).to(device)
    s = torch.randn(1, 3, 256, 256, requires_grad=True).to(device)

    mask_c = torch.randn(1, 2, 256, 256, requires_grad=True).to(device)
    mask_s = torch.randn(1, 2, 256, 256, requires_grad=True).to(device)

    diff_c = torch.randn(1, 136, 256, 256, requires_grad=True).to(device)
    diff_s = torch.randn(1, 136, 256, 256, requires_grad=True).to(device)

    lms_c = torch.randn(1, 68, 2, requires_grad=True).to(device)
    lms_s = torch.randn(1, 68, 2, requires_grad=True).to(device)

    input = ['c', 's', 'mask_c', 'mask_s', 'diff_c', 'diff_s', 'lms_c', 'lms_s']
    output = ['output']

    torch.onnx.export(self.solver.G, (c, s, mask_c, mask_s, diff_c, diff_s, lms_c, lms_s), 'model.onnx', verbose=True, 
    opset_version=15, input_names=input, output_names=output)

Cannot convert to onnx using the above code: Exporting the operator inverse to ONNX opset version 15 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub.

Hope you can teach me how to solve the problem.

I don't know about the code but this error "Exporting the operator inverse to ONNX opset version 15 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub " i think can be fixed simply by using pytorch nightly

[ZTMIDGO]

@Fcucgvhhhvjv Based on the support list listed by pytorch https://pytorch.org/docs/stable/onnx_supported_aten_ops.html, perhaps only replacing the operator will solve the problem