Home Forums Model Development Kit (MDK) Caffe.proto file merge confusion

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #1986
    Francisco Soto
    Participant

    Hi,
    Do you have an caffe.proto already merged? I think it is kind of difficult to merge line by line, when there is more than 1K lines. Do you hub a docker of this to deploy it faster?

    Thanks,
    Sebatián.

    #1989
    Arpine SoghoyanArpine Soghoyan
    Moderator

    You only need to merge, if using Caffe version other than v1.0. Which Caffe version are you using ?

    #1994
    Francisco Soto
    Participant

    I’m using the last Caffe version, I think is a little more improved than the 1.0 one. But it is not in the releases. But, how can you “merge” the file? Do you mean, copy the whole file and put it on the last part of the original file? Because by “merge” I have a confusion, reading each of 1k line in a file is a little bit difficult. Thanks.

    #1996
    Arpine SoghoyanArpine Soghoyan
    Moderator

    The easiest way would be using ultraedit compare files option. It’ll highlight the differences and you can make the changes accordingly. Appending the file would be wrong as some of the modifications are within functions.

    #2005
    Francisco Soto
    Participant

    Finally, I compiled caffe, but cannot execute MDK example, like make gnet1, everything ends up in a segmentation fault. Can you provide some github fork of the project to download it and run it without problems? A fork that can run “make gnet1” without problem in CPU for Ubuntu 16.04 LTS? Thanks.

    #2006
    Francisco Soto
    Participant

    Or at least, provide an instruction clear, because “merge” is to subjective, at least, instructions to download Caffe and “replace” whole files to compile it without problems (Makefile.config, CMakeList.txt and coffe.proto). We’d appreciate that. Cannot even run the first examples and try a lot of staff before write this feed. Thanks.

    #2011
    Arpine SoghoyanArpine Soghoyan
    Moderator

    In our next version of the caffe MDK tool we actually made improvements, and will definitely consider your recommendations. Can you please provide the error log ? you can directly send me an email.

    #2030
    Francisco Soto
    Participant

    gticoefconvt: 3.1-p000: (c) Copyright 2017-2018 Gyrfalcon Technology Inc.
    Input :
    Network : inputs/gnet1/gnet1.prototxt
    Weights : inputs/gnet1/gnet1.caffemodel
    JsonConfig: inputs/gnet1/gnet1.json
    Test Image: inputs/gnet1/gnet1_test_image.dat
    updateJson: 1
    meanRemove: 0
    slicing : 0
    batchNorm : 0
    Output:
    EditedWeight : cnn_weights_gnet1/Gain_Edited.caffemodel
    jsonConfEdited : cnn_weights_gnet1/gtiEditedConfig.json
    Output COEF : cnn_weights_gnet1/output_coef_conv.dat

    [libprotobuf WARNING google/protobuf/io/coded_stream.cc:537] Reading dangerously large protocol message. If the message turns out to be larger than 2147483647 bytes, parsing will be halted for security reasons. To increase the limit (or to disable these warnings), see CodedInputStream::SetTotalBytesLimit() in google/protobuf/io/coded_stream.h.
    [libprotobuf WARNING google/protobuf/io/coded_stream.cc:78] The total number of bytes read was 553432486
    init_caffe(): End of loading network inputs/gnet1/gnet1.prototxt, weight: inputs/gnet1/gnet1.caffemodel …
    Segmentation fault (core dumped)

    #2031
    Francisco Soto
    Participant

    This is a full error log when “make all”

    gticoefconvt: 3.1-p000: (c) Copyright 2017-2018 Gyrfalcon Technology Inc.
    Input :
    Network : inputs/gnet1/gnet1.prototxt
    Weights : inputs/gnet1/gnet1.caffemodel
    JsonConfig: inputs/gnet1/gnet1.json
    Test Image: inputs/gnet1/gnet1_test_image.dat
    updateJson: 1
    meanRemove: 0
    slicing : 0
    batchNorm : 0
    Output:
    EditedWeight : cnn_weights_gnet1/Gain_Edited.caffemodel
    jsonConfEdited : cnn_weights_gnet1/gtiEditedConfig.json
    Output COEF : cnn_weights_gnet1/output_coef_conv.dat

    [libprotobuf WARNING google/protobuf/io/coded_stream.cc:537] Reading dangerously large protocol message. If the message turns out to be larger than 2147483647 bytes, parsing will be halted for security reasons. To increase the limit (or to disable these warnings), see CodedInputStream::SetTotalBytesLimit() in google/protobuf/io/coded_stream.h.
    [libprotobuf WARNING google/protobuf/io/coded_stream.cc:78] The total number of bytes read was 553432486
    init_caffe(): End of loading network inputs/gnet1/gnet1.prototxt, weight: inputs/gnet1/gnet1.caffemodel …
    Segmentation fault (core dumped)
    /home/tiantg/caffe-master
    gticoefconvt: 3.1-p000: (c) Copyright 2017-2018 Gyrfalcon Technology Inc.
    Input :
    Network : inputs/sres/sres.prototxt
    Weights : inputs/sres/sres.caffemodel
    JsonConfig: inputs/sres/sres.json
    Test Image: inputs/sres/sres_test_image.dat
    updateJson: 1
    meanRemove: 0
    slicing : 0
    batchNorm : 0
    Output:
    EditedWeight : cnn_weights_sres/Gain_Edited.caffemodel
    jsonConfEdited : cnn_weights_sres/gtiEditedConfig.json
    Output COEF : cnn_weights_sres/output_coef_conv.dat

    init_caffe(): End of loading network inputs/sres/sres.prototxt, weight: inputs/sres/sres.caffemodel …
    Segmentation fault (core dumped)
    echo “LD_LIBRARY_PATH = /home/tiantg/caffe-master/build/lib”
    LD_LIBRARY_PATH = /home/tiantg/caffe-master/build/lib
    ./utils/gtimodelconvt inputs/style/style.prototxt inputs/style/style.caffemodel inputs/style/style.json inputs/style/style_imgmean.txt inputs/style/style_test_image.dat cnn_weights_style 1 1 2>&1 | tee log/log_style_debug.txt
    gticoefconvt: 3.1-p000: (c) Copyright 2017-2018 Gyrfalcon Technology Inc.
    Input :
    Network : inputs/style/style.prototxt
    Weights : inputs/style/style.caffemodel
    JsonConfig: inputs/style/style.json
    Test Image: inputs/style/style_test_image.dat
    updateJson: 1
    meanRemove: 0
    slicing : 0
    batchNorm : 0
    Output:
    EditedWeight : cnn_weights_style/Gain_Edited.caffemodel
    jsonConfEdited : cnn_weights_style/gtiEditedConfig.json
    Output COEF : cnn_weights_style/output_coef_conv.dat

    init_caffe(): End of loading network inputs/style/style.prototxt, weight: inputs/style/style.caffemodel …
    Segmentation fault (core dumped)
    echo “./utils/gtimodelconvt inputs/gnet18svic/gnet18.prototxt inputs/gnet18svic/gnet18svic.caffemodel inputs/gnet18svic/network_Gnet18_gain.json inputs/gnet18svic/test_image_peppers2.dat cnn_weights_gnet18svic ” | tee log/log_gnet18_debug.txt
    ./utils/gtimodelconvt inputs/gnet18svic/gnet18.prototxt inputs/gnet18svic/gnet18svic.caffemodel inputs/gnet18svic/network_Gnet18_gain.json inputs/gnet18svic/test_image_peppers2.dat cnn_weights_gnet18svic
    echo “./utils/gtimodelconvt option <DEBUGMODE> <Json Updated> <ImageMean> <Slicing> <MeanRemove> <BatchNormMerge>”
    ./utils/gtimodelconvt option <DEBUGMODE> <Json Updated> <ImageMean> <Slicing> <MeanRemove> <BatchNormMerge>
    ./utils/gtimodelconvt inputs/gnet18svic/gnet18.prototxt inputs/gnet18svic/gnet18svic.caffemodel inputs/gnet18svic/network_Gnet18_gain.json inputs/gnet18svic/gnet18svic_imgmean.txt inputs/gnet18svic/test_image_peppers2.dat cnn_weights_gnet18svic 3 1 1 1 1 2>&1 | tee -a log/log_gnet18_debug.txt
    gticoefconvt: 3.1-p000: (c) Copyright 2017-2018 Gyrfalcon Technology Inc.
    Input :
    Network : inputs/gnet18svic/gnet18.prototxt
    Weights : inputs/gnet18svic/gnet18svic.caffemodel
    JsonConfig: inputs/gnet18svic/network_Gnet18_gain.json
    Test Image: inputs/gnet18svic/test_image_peppers2.dat
    updateJson: 1
    meanRemove: 1
    slicing : 1
    batchNorm : 1
    Output:
    EditedWeight : cnn_weights_gnet18svic/Gain_Edited.caffemodel
    jsonConfEdited : cnn_weights_gnet18svic/gtiEditedConfig.json
    Output COEF : cnn_weights_gnet18svic/output_coef_conv.dat

    init_caffe(): End of loading network inputs/gnet18svic/gnet18.prototxt, weight: inputs/gnet18svic/gnet18svic.caffemodel …
    gtimodelconvt: /usr/include/boost/smart_ptr/shared_ptr.hpp:648: typename boost::detail::sp_member_access<T>::type boost::shared_ptr<T>::operator->() const [with T = caffe::Layer<float>; typename boost::detail::sp_member_access<T>::type = caffe::Layer<float>*]: Assertion `px != 0′ failed.
    convLayerNum = 0, layercount = 18
    Aborted (core dumped)

Viewing 9 posts - 1 through 9 (of 9 total)
  • You must be logged in to reply to this topic.