|
用Rknn_toolkit 1.6.0版本转一个Caffe模型(用在RK3399Pro上面),有如下报错:
---> Load model
W Unhandle status: the input shape of reshape layer split_28_split2part_0_reshape_reshape_28 is not 4-D
E Catch exception when loading caffe model: ./resnest50.prototxt!
E Traceback (most recent call last):
E File "rknn/api/rknn_base.py", line 247, in rknn.api.rknn_base.RKNNBase.load_caffe
E File "rknn/base/RKNNlib/RK_nn.py", line 74, in rknn.base.RKNNlib.RK_nn.RKnn.load_caffe
E File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 323, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build
E File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 346, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 346, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 346, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E [Previous line repeated 43 more times]
E File "rknn/base/RKNNlib/RKNNnetbuilder.py", line 353, in rknn.base.RKNNlib.RKNNnetbuilder.RKNNNetBuilder.build_layer
E File "rknn/base/RKNNlib/layer/RKNNlayer.py", line 265, in rknn.base.RKNNlib.layer.RKNNlayer.RKNNLayer.compute_shape
E File "rknn/base/RKNNlib/layer/tile.py", line 19, in rknn.base.RKNNlib.layer.tile.Tile.compute_out_shape
E IndexError: list index out of range
问题是同样的模型用Rknn_toolkit 1.4.0的版本是能够转换成功的,请问在1.6.0版本上对这个报错有什么解决办法吗?
|
|