search

microsoft / MMdnn

MMdnn is a set of tools to help users inter-operate among different deep learning frameworks. E.g. model conversion and visualization. Convert models between Caffe, Keras, MXNet, Tensorflow, CNTK, PyTorch Onnx and CoreML.
MIT License
5.79k stars 964 forks source link

mobilenetv2 tf to caffe #592

Open xubin19939 opened 5 years ago

xubin19939 commented 5 years ago

Platform (like ubuntu 16.04/win10):ubuntu 16.04

Python version:2.7.12

Source framework with version (like Tensorflow 1.4.1 with GPU):Tensorflow 1.8 with GPU

Destination framework with version (like CNTK 2.3 with GPU):

Pre-trained model path (webpath or webdisk path):https://github.com/xubin19939/models_mobilenet2 I am confused if mmdnn support biasadd or not? the issue occured in slim.separable_conv2d

Running scripts: tensorflow -in modelmodel.ckpt.meta -iw modelmodel.ckpt --dstNodeName Softmax -df caffe -om tf_resnet Parse file [modelmodel.ckpt.meta] with binary format successfully. Tensorflow model file [modelmodel.ckpt.meta] loaded successfully. Tensorflow checkpoint file [modelmodel.ckpt] loaded successfully. [56] variables loaded. Tensorflow has not supported operator [BiasAdd] with name [n1_26/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n1_256a/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n1_28a1/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n1_28a2/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n3_14a1/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n3_14a2/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n3_14a3/depthwise/BiasAdd]. Tensorflow has not supported operator [BiasAdd] with name [n4_112a/depthwise/BiasAdd]. Tensorflow has not supported operator [Softmax] with name [Softmax]. IR network structure is saved as [3733adaf1a094712a6ab962ec02e6ead.json]. IR network structure is saved as [3733adaf1a094712a6ab962ec02e6ead.pb]. IR weights are saved as [3733adaf1a094712a6ab962ec02e6ead.npy]. Traceback (most recent call last): File "/usr/local/bin/mmconvert", line 11, in sys.exit(_main()) File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/_script/convert.py", line 108, in _main ret = IRToCode._convert(code_args) File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/_script/IRToCode.py", line 9, in _convert from mmdnn.conversion.caffe.caffe_emitter import CaffeEmitter File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/caffe/caffe_emitter.py", line 12, in import caffe File "/data/caffe-master/python/caffe/init.py", line 1, in from .pycaffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, RMSPropSolver, AdaDeltaSolver, AdamSolver, NCCL, Timer File "/data/caffe-master/python/caffe/pycaffe.py", line 13, in from ._caffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, \ ImportError: dynamic module does not define module export function (PyInit__caffe)

xubin19939 commented 5 years ago

try to convert mobilenet tf model to caffe from frozen graph,get these errors: I0222 16:02:58.296536 1349 net.cpp:84] Creating Layer n1_26_depthwise_BiasAdd I0222 16:02:58.296542 1349 net.cpp:406] n1_26_depthwise_BiasAdd <- n1_26_depthwise_depthwise I0222 16:02:58.296551 1349 net.cpp:380] n1_26_depthwise_BiasAdd -> n1_26_depthwise_BiasAdd F0222 16:02:58.296568 1349 layer.hpp:354] Check failed: MinBottomBlobs() <= bottom.size() (2 vs. 1) Eltwise Layer takes at least 2 bottom blob(s) as input. Check failure stack trace: Aborted (core dumped)