转efficientnet.tflite到rknn时出现异常,'NoneType' object has no attribu...

D Convert layer conv_2d
W Rename efficientnet-lite0/model/head/Relu6;efficientnet-lite0/model/head/tpu_batch_normalization/FusedBatchNormV3;efficientnet-lite0/head/conv2d/kernel;efficientnet-lite0/model/head/conv2d to rknn_efficientnet-lite0/model/head/Relu6;efficientnet-lite0/model/head/tpu_batch_normalization/FusedBatchNormV3;efficientnet-lite0/head/conv2d/kernel;efficientnet-lite0/model/head/conv2d
W Rename efficientnet-lite0/model/head/Relu6;efficientnet-lite0/model/head/tpu_batch_normalization/FusedBatchNormV3;efficientnet-lite0/head/conv2d/kernel;efficientnet-lite0/model/head/conv2d to rknn_efficientnet-lite0/model/head/Relu6;efficientnet-lite0/model/head/tpu_batch_normalization/FusedBatchNormV3;efficientnet-lite0/head/conv2d/kernel;efficientnet-lite0/model/head/conv2d
D Convert layer average_pool_2d
D Convert layer reshape
E Catch exception when loading tflite model: ./efficientnet.tflite!
E Traceback (most recent call last):
E   File "rknn/api/rknn_base.py", line 365, in rknn.api.rknn_base.RKNNBase.load_tflite
E   File "rknn/base/RKNNlib/converter/tflite_loader.py", line 75, in rknn.base.RKNNlib.converter.tflite_loader.TFLiteLoader.load
E   File "rknn/base/RKNNlib/converter/tflite_loader.py", line 393, in rknn.base.RKNNlib.converter.tflite_loader.ModelParser.parse
E   File "rknn/base/RKNNlib/converter/tflite_loader.py", line 339, in rknn.base.RKNNlib.converter.tflite_loader.ModelParser._build_layers
E   File "rknn/base/RKNNlib/converter/tflite_loader.py", line 235, in rknn.base.RKNNlib.converter.tflite_loader.ModelParser._get_layer_RKNN_op
E   File "rknn/base/RKNNlib/converter/lite/RKNN_map_op.py", line 505, in rknn.base.RKNNlib.converter.lite.RKNN_map_op.RKNNMapOp.map_RKNN_ops
E   File "rknn/base/RKNNlib/converter/lite/RKNN_map_op.py", line 299, in rknn.base.RKNNlib.converter.lite.RKNN_map_op._reshape
E AttributeError: 'NoneType' object has no attribute 'NewShapeAsNumpy'

jefferyzhang 发表于 2020-3-31 23:39
请将原始模型和转换脚本一起上传到百度网盘，然后链接发给我们，我提给NPU部门调试。还有请将以下信息提供 ...

链接：https://pan.baidu.com/s/17ZOc9E4nfheQqTEWcE8VTQ
提取码：weh1

TensorFlow-GPU 1.14
rknn_toolkit 1.3.0

jefferyzhang 发表于 2020-4-1 14:25
好的，问题已经提交了。
你的tf可以降到1.13.x试下，1.14会遇到一些op问题

试了tf 1.13.1的, 报了同样的错误

hanrb 发表于 2020-4-1 20:40
试了tf 1.13.1的, 报了同样的错误

有更新了吗

jefferyzhang 发表于 2020-4-10 08:49
NPU部门回复：

转换rknn失败的bug我们这边已经解决了，efficientnet-lite0-fp32.tflite模型在tflite runt ...

好的,谢谢