onnx 转换失败

[thomaszheng]

模型来源： onnx 模型说明： 文本检测dbnet网络 模型文件：链接: https://pan.baidu.com/s/1ml6PFUdNey2TnHjSOcCqmg 密码: pieo

x2paddle --framework=onnx --model=ocr_detector.onnx --save_dir=pd_model

转换过程出错提示如下： python 3.6.2 x2paddle 1.3.1

paddle.__version__ = 2.0.0
Now translating model from onnx to paddle.
model ir_version: 6, op version: 9
shape inferencing ...
Stopping at incomplete shape inference at Concat: x2paddle_312
node inputs:
name: "x2paddle_296"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
        dim_value: 1
      }
      dim {
        dim_value: 96
      }
      dim {
        dim_value: 64
      }
      dim {
        dim_value: 64
      }
    }
  }
}

name: "x2paddle_299"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
        dim_value: 1
      }
      dim {
        dim_value: 96
      }
      dim {
        dim_value: 64
      }
      dim {
        dim_value: 64
      }
    }
  }
}

name: "x2paddle_302"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
        dim_value: 1
      }
      dim {
        dim_value: 96
      }
      dim {
        dim_value: 64
      }
      dim {
        dim_value: 64
      }
    }
  }
}

name: "x2paddle_305"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
        dim_value: 1
      }
      dim {
        dim_value: 96
      }
      dim {
        dim_value: 64
      }
      dim {
        dim_value: 64
      }
    }
  }
}

name: "x2paddle_311"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
        dim_param: "Upsample93_o0__d0"
      }
      dim {
        dim_param: "Upsample93_o0__d1"
      }
      dim {
        dim_param: "Upsample93_o0__d2"
      }
      dim {
        dim_param: "Upsample93_o0__d3"
      }
    }
  }
}

node outputs:
name: "x2paddle_312"
type {
  tensor_type {
    elem_type: 1
    shape {
      dim {
      }
      dim {
        dim_param: "Upsample93_o0__d1 + 384"
      }
      dim {
      }
      dim {
      }
    }
  }
}

!!!!!!!!!!
shape inferenced.
Now, onnx2paddle support convert onnx model opset_verison [9],opset_verison of your onnx model is 9, automatically treated as op_set: 9.
Total nodes: 119
Nodes converting ...
Converting node 299 ...
Nodes converted

[yeliang2258]

模型能够转换成功，你可以先尝试看转换的模型能否正确部署。

[thomaszheng]

是说这个报错信息是不影响的，模型是成功转换的

[yeliang2258]

说这个报错信息是不影响的，模型是成功转换的

转换成功的模型试试可以用吗？

[thomaszheng]

确实可以推理，这个错误目前看不影响转换后的模型使用