Skip to content

Latest commit

 

History

History
228 lines (156 loc) · 10 KB

installation.md

File metadata and controls

228 lines (156 loc) · 10 KB
Raw

1. User Quick Guide

GitHub license

This Installation will help you get started to setup TensorFlow-HRT on RK3399 quickly.

2. Preparation on build server

2.1 Update apt source list

  sudo dpkg --add-architecture arm64
  
  sudo mv /etc/apt/sources.list /etc/apt/sources.list.bak

create file "/etc/apt/sources.list.d/tsinghua.list" with following content

  deb [arch=amd64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ xenial main restricted universe multiverse
  deb [arch=amd64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ xenial-updates main restricted universe multiverse
  deb [arch=amd64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ xenial-security main restricted universe multiverse
  deb [arch=amd64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ xenial-backports main restricted universe multiverse
  deb [arch=arm64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial main restricted universe multiverse
  deb [arch=arm64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial-updates main restricted universe multiverse
  deb [arch=arm64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial-security main restricted universe multiverse
  deb [arch=arm64] https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial-backports main restricted universe multiverse

2.2 General dependencies installation

sudo apt-get update

sudo apt-get install git scons gcc-aarch64-linux-gnu g++-aarch64-linux-gnu scons rsync

sudo apt-get install python-numpy python-dev python-pip python-wheel

sudo apt-get install libpython2.7-dev:arm64

2.3 Install bazel (release 0.7.0 or above)

TensorFlow uses bazel to build. If bazel is not installed on your system, install it now by following guide at https://docs.bazel.build/versions/master/install.html.

2.2 Download source code

cd ~

Download "ACL"

git clone https://github.com/ARM-software/ComputeLibrary.git

Download "TensorFlow-HRT" :

git clone https://github.com/OAID/TensorFlow-HRT

3. Build TensorFlow-HRT

3.1 Build ACL :

cd ~/ComputeLibrary
git checkout v17.12
scons Werror=1 -j4 debug=0 asserts=1 neon=1 opencl=1 embed_kernels=1 os=linux arch=arm64-v8a
scp build/*.so firefly@192.168.3.211:/home/firefly/

3.2 Build TensorFlow-HRT

*1 Update cross build setting for ACL Modify following 3 lines in “TensorFlow-HRT/tools/aarch64_compiler/CROSSTOOL” to ACL path in build server.

35: linker_flag: "-L/home/[user name]/ComputeLibrary/build"
36: linker_flag: "-L/home/[user name]/ComputeLibrary/build/opencl-1.2-stubs/"
44: cxx_builtin_include_directory: "/home/[user name]/ComputeLibrary"

*2 Run TensorFlow-HRT configure

cd ~/TensorFlow-HRT
./configure

  WARNING: Running Bazel server needs to be killed, because the startup options are different.
  You have bazel 0.7.0 installed.
  Please specify the location of python. [Default is /usr/bin/python]: 
  Found possible Python library paths:
    /usr/local/lib/python2.7/dist-packages
    /usr/lib/python2.7/dist-packages
  Please input the desired Python library path to use.  Default is [/usr/local/lib/python2.7/dist-packages]

  Do you wish to build TensorFlow with jemalloc as malloc support? [Y/n]: 
  jemalloc as malloc support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with Google Cloud Platform support? [Y/n]: n
  No Google Cloud Platform support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with Hadoop File System support? [Y/n]: n
  No Hadoop File System support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with Amazon S3 File System support? [Y/n]: n
  No Amazon S3 File System support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with XLA JIT support? [y/N]: n
  No XLA JIT support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with GDR support? [y/N]: n
  No GDR support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with VERBS support? [y/N]: n
  No VERBS support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with OpenCL SYCL support? [y/N]: n
  No OpenCL SYCL support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with CUDA support? [y/N]: n
  No CUDA support will be enabled for TensorFlow.

  Do you wish to build TensorFlow with MPI support? [y/N]: n
  No MPI support will be enabled for TensorFlow.

  Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]: 

  Add "--config=mkl" to your bazel command to build with MKL support.

  Please note that MKL on MacOS or windows is still not supported. If you would like to use a local MKL instead of downloading, 
  please set the environment variable "TF_MKL_ROOT" every time before build.
  Configuration finished

*3 build pip package (need to change ComputeLibrary path)

cd ~/TensorFlow-HRT

bazel build -c opt --cxxopt=-fexceptions --copt="-I/home/[user name]/ComputeLibrary/" --copt="-I/home/[user name]/ComputeLibrary/include" \
  --cpu=aarch64 --crosstool_top=//tools/aarch64_compiler:toolchain --host_crosstool_top=@bazel_tools//tools/cpp:toolchain --host_copt="-DUSE_ACL=1" \
  --copt="-DUSE_ACL=1" --copt="-DTEST_ACL=1" --copt="-DUSE_PROFILING=1" --incompatible_load_argument_is_label=false --verbose_failures \
  //tensorflow/tools/pip_package:build_pip_package

bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/install/

mv /tmp/install/tensorflow-1.4.0-cp27-cp27mu-linux_x86_64.whl tensorflow-1.4.0-cp27-cp27mu-linux_aarch64.whl

If there is a following build error

  /home/xxxx/.cache/bazel/_bazel_xxxx/e744db622218e02ee1e2cbf3b8750f17/external/nsync/BUILD:402:13: Configurable attribute "copts" 
  doesn't match this configuration (would a default condition help?).
  Conditions checked:

  @nsync//:android_arm
  ...
  @nsync//:msvc_windows_x86_64.

Open “/home/xxx/.cache/bazel/_bazel_xxxx/e744db622218e02ee1e2cbf3b8750f17/external/nsync/BUILD”, Add a default conditions there as marked as red line below:

  NSYNC_OPTS_GENERIC = select({
    # Select the CPU architecture include directory.
    # This select() has no real effect in the C++11 build, but satisfies a
    # #include that would otherwise need a #if.
    ":gcc_linux_x86_64_1": ["-I" + pkg_path_name() + "/platform/x86_64"],
    ":gcc_linux_x86_64_2": ["-I" + pkg_path_name() + "/platform/x86_64"],
    ":gcc_linux_aarch64": ["-I" + pkg_path_name() + "/platform/aarch64"],
    ":gcc_linux_ppc64": ["-I" + pkg_path_name() + "/platform/ppc64"],
    ":clang_macos_x86_64": ["-I" + pkg_path_name() + "/platform/x86_64"],
    ":ios_x86_64": ["-I" + pkg_path_name() + "/platform/x86_64"],
    ":android_x86_32": ["-I" + pkg_path_name() + "/platform/x86_32"],
    ":android_x86_64": ["-I" + pkg_path_name() + "/platform/x86_64"],
    ":android_armeabi": ["-I" + pkg_path_name() + "/platform/arm"],
    ":android_arm": ["-I" + pkg_path_name() + "/platform/arm"],
    ":android_arm64": ["-I" + pkg_path_name() + "/platform/aarch64"],
    ":msvc_windows_x86_64": ["-I" + pkg_path_name() + "/platform/x86_64"],
    "//conditions:default": [],
  }) + [

*4 Build label_image

  cd ~/TensorFlow-HRT
  
  bazel build -c opt --incompatible_load_argument_is_label=false --cxxopt=-fexceptions --copt="-I/home/[user name]/ComputeLibrary/" \
    --copt="-I/home/[user name]/ComputeLibrary/include" --cpu=aarch64 --crosstool_top=//tools/aarch64_compiler:toolchain \
    --host_crosstool_top=@bazel_tools//tools/cpp:toolchain --copt="-DUSE_ACL=1" --copt="-DTEST_ACL=1" --copt="-DUSE_PROFILING=1" --verbose_failures \
    //tensorflow/examples/label_image/...

4. Run tests

4.1 copy and install build files

  ssh firefly@192.168.3.211 "mkdir -p test/tensorflow/examples/label_image/data/" 
  
  cd ~/TensorFlow-HRT 
  
  scp acl_openailab/test/* firefly@192.168.3.211:/home/firefly/test/
  
  scp tensorflow-1.4.0-cp27-cp27mu-linux_aarch64.whl firefly@192.168.3.211:/home/firefly/test/
  
  scp tensorflow/python/kernel_tests/acl*.py firefly@192.168.3.211:/home/firefly/test/
  
  scp bazel-bin/tensorflow/examples/label_image/label_image firefly@192.168.3.211:/home/firefly/test/
  
  scp tensorflow/examples/label_image/data/grace_hopper.jpg firefly@192.168.3.211:/home/firefly/test/tensorflow/examples/label_image/data/
  
  ssh firefly@192.168.3.211
  
  cd test
  
  sudo pip install tensorflow-1.4.0-cp27-cp27mu-linux_aarch64.whl
  
  7z x models.7z.001

4.2 Run label_image Example

  ssh firefly@192.168.3.211
  
  cd test
  
  tar -C tensorflow/examples/label_image/data -xvzf inception_v3_2016_08_28_frozen.pb.tar.gz
  
  LD_LIBRARY_PATH=/usr/local/lib/python2.7/dist-packages/tensorflow/ ./label_image

output message

  2018-02-09 05:58:38.309086: I tensorflow/examples/label_image/main.cc:250] military uniform (653): 0.834305
  2018-02-09 05:58:38.309265: I tensorflow/examples/label_image/main.cc:250] mortarboard (668): 0.0218695
  2018-02-09 05:58:38.309299: I tensorflow/examples/label_image/main.cc:250] academic gown (401): 0.0103581
  2018-02-09 05:58:38.309329: I tensorflow/examples/label_image/main.cc:250] pickelhaube (716): 0.00800819
  2018-02-09 05:58:38.309359: I tensorflow/examples/label_image/main.cc:250] bulletproof vest (466): 0.00535092

4.3 Run Unit test

  ssh firefly@192.168.3.211
  
  cd test
  
  ls acl*.py | xargs -I {} python {}

output message

testInceptionFwd_0 [4, 5, 5, 124] [1, 1, 124, 12] [4, 5, 5, 12] 1
testInceptionFwd_1 [4, 8, 8, 38] [1, 1, 38, 38] [4, 8, 8, 38] 1
testInceptionFwd_2 [4, 8, 8, 38] [1, 1, 38, 38] [4, 8, 8, 38] 1
...
Testing InceptionFwd %s ([4, 8, 8, 176], [1, 1, 176, 19], 1, 'SAME')
Testing InceptionFwd %s ([4, 8, 8, 176], [1, 1, 176, 19], 1, 'SAME')
......
----------------------------------------------------------------------
Ran 62 tests in 2.735s

OK