请问一下这个是因为rknn中没有这个op操作么

[复制链接] · 发表于 2019-8-20 14:26:37

将yolo v3——mobilenet v3的keras h5 权重转到rknn失败，显示‘'NoneType' object is not iterable’。又先换成了tflite再转还是失败，下面是tflite转rknn的错误代码，能帮我看下是哪里出了问题么。。。

--> Loading model
done
--> Building model
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
W Custom op maximum
E Catch exception when building RKNN model!
T Traceback (most recent call last):
T File "rknn/api/rknn_base.py", line 521, in rknn.api.rknn_base.RKNNBase.build
T File "rknn/api/rknn_base.py", line 440, in rknn.api.rknn_base.RKNNBase._build
T File "rknn/base/ovxconfiggenerator.py", line 197, in rknn.base.ovxconfiggenerator.generate_vx_config_from_files
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 382, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator.generate
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 339, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._gen_unify_case
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 241, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._generate_ovxlib_case
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/vxnetgenerator.py", line 226, in rknn.base.RKNNlib.app.exporter.ovxlib_case.vxnetgenerator.VXNetGenerator.save
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/vxnode_mapper.py", line 140, in rknn.base.RKNNlib.app.exporter.ovxlib_case.vxnode_mapper.VXNodeMapper.save_source
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/vxnetclientgenerator.py", line 193, in rknn.base.RKNNlib.app.exporter.ovxlib_case.vxnetclientgenerator.ClientGenerator.save_source
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/vxnetclientgenerator.py", line 178, in rknn.base.RKNNlib.app.exporter.ovxlib_case.vxnetclientgenerator.ClientGenerator.load_source
T File "rknn/base/RKNNlib/app/exporter/ovxlib_case/vxnetclientgenerator.py", line 156, in rknn.base.RKNNlib.app.exporter.ovxlib_case.vxnetclientgenerator.load
T FileNotFoundError: [Errno 2] No such file or directory: '/home/abakus/下载/xunlian/scratches/keras-yolo3-master/vxcode/template/client/rknn_#NETWORK_LOWER#_client_tab.c'
done
E RKNN model data is None, please load model first!

只看该作者 · 发表于 2019-8-20 15:58:50

hisping 发表于 2019-8-20 15:15
rknn-toolkit中load_darknet支持直接加载yolov3模型，为什么你要转成keras h5格式？另外rknn-toolkit不支持 ...

我之前重新训练过mobilenet v3框架的yolov3，使用keras实现的，所以就是h5格式的。我是先转的pb格式。pb转rknn时load_tensorflow出现'NoneType' object is not iterable“这个错误，pb转成tflite后再转rknn就是上面那一段错误。。