onnx转rknn问题

D RKNN output shape(convolution): (1 10 10 768)
D Process Relu_57_19 ...
D RKNN output shape(relu): (1 10 10 768)
D Process Conv_58_18 ...
D RKNN output shape(convolution): (1 8 8 768)
D Process Relu_60_15 ...
D RKNN output shape(relu): (1 8 8 768)
D Process Conv_61_14 ...
D RKNN output shape(convolution): (1 6 6 512)
D Process Conv_63_16 ...
D RKNN output shape(convolution): (1 4 4 10240)
D Process Reshape_75_12 ...
D RKNN output shape(reshape): (20 4 4 512)
D Process permute_8 ...
D RKNN output shape(permute): (4 4 512 20)
D Process Conv_102_7 ...
E Catch exception when loading onnx model: siamrpn.onnx!
E Traceback (most recent call last):
E   File "rknn/api/rknn_base.py", line 519, in rknn.api.rknn_base.RKNNBase.load_onnx
E   File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 286, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build
E   File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 305, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E   File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 305, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E   File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 311, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E   File "rknn/base/RKNNlib/layer/RKNNlayer.py", line 262, in rknn.base.RKNNlib.layer.RKNNlayer.RKNNLayer.compute_shape
E   File "rknn/base/RKNNlib/layer/conv2d_op.py", line 91, in rknn.base.RKNNlib.layer.conv2d_op.Conv2DOP.compute_out_shape
E AttributeError: 'LayerParams' object has no attribute 'weights'
Load siamrpn failed!

环境：rknn-toolkit1.3.2 pytorch1.2 onnx1.4.1
Conv_102这一层的权重不是训练得到的，是计算出来的，不同输入这一层权重不同，是否是这个问题？
onnx文件本身可以加载没问题。