Help with reshape difference between tensorflow and tensorrt

When I run my model, I get this error:
[TensorRT] ERROR: inference/coefficients/fusion/add: elementwise inputs must have same dimensions or follow broadcast rules (input dimensions were [64,16,16] and [1,1,64])
[TensorRT] ERROR: inference/coefficients/prediction/conv1/Conv2D: at least three non-batch dimensions are required for input
[TensorRT] ERROR: UFFParser: Parser error: inference/coefficients/prediction/conv1/BiasAdd: The input to the Scale Layer is required to have a minimum of 3 dimensions.
[TensorRT] ERROR: Network must have at least one output
According the error logs, I found that the reshape op before “inference/coefficients/fusion/add” in uff model give different result in pb model. And I think this case this error.
And here is my inference code, model and network shown in tensorboard:
can you help me to solve this problem?