使用变化检测模型CDNet进行解译时报错

[mj-moose]

基于paddlers训练的CDNet模型导出至Geoview后进行解译，报错内容如下：

File "D:\GeoView-release-0.1.1\backend\applications\interface\change_detection1.py", line 57, in
execute(model_path, data_path, out_dir, window_size=256, stride=128)
File "D:\GeoView-release-0.1.1\backend\applications\interface\change_detection1.py", line 25, in execute
predictor.slider_predict(
File "d:\geoview\paddlers\paddlers\deploy\predictor.py", line 336, in slider_predict
slider_predict(
File "d:\geoview\paddlers\paddlers\tasks\utils\slider_predict.py", line 512, in slider_predict
batch_out = predict_func(batch_data, transforms=transforms)
File "d:\geoview\paddlers\paddlers\deploy\predictor.py", line 281, in predict
results = self._run(images=images, topk=topk, transforms=transforms)
File "d:\geoview\paddlers\paddlers\deploy\predictor.py", line 226, in _run
net_outputs = self.raw_predict(preprocessed_input)
File "d:\geoview\paddlers\paddlers\deploy\predictor.py", line 210, in raw_predict
self.predictor.run()
ValueError: In user code:
........
InvalidArgumentError: index should less than output tensor height * output tensor width. Expected 4 < 4, but got 4 >= 4. Please check input value.
[Hint: Expected index < output_feasize, but received index:4 >= output_feasize:4.] (at ..\paddle\phi\kernels\cpu\unpool_kernel.cc:65)
[operator < unpool > #error]