Skip to content

Latest commit

 

History

History
104 lines (74 loc) · 2.75 KB

install.md

File metadata and controls

104 lines (74 loc) · 2.75 KB
Raw

Installation Guide

This guide gives instructions on how to build and test Tengine on your system.

1. Preparation

1.1 Download source code

To get started, git clone the latest Tengine repository.

git clone https://github.com/OAID/tengine/

1.2 Install Depency Libraries

  • libprotobuf: for load caffemodel

     sudo apt install libprotobuf-dev protobuf-compiler libboost-all-dev libgoogle-glog-dev

  • libopencv: for image preprocessing in test samples

     sudo apt install libopencv-dev

  • Caffe (Optional): use Caffe's operators for verifing your operator implementations in Tengine

    Please see http://caffe.berkeleyvision.org/installation.html

1.3 Prepare config files

  • copy config example file 
     cd ~/tengine
 
 cp makefile.config.example makefile.config
  • edit makefile.config

    • arm64

      By default, CONFIG_ARCH_ARM64 option is valid.

    • if your want to use Caffe, set

       CONFIG_CAFFE_REF=y
 CAFFE_ROOT = /home/firefly/caffe (your caffe path)

    • if you want to run using Openblas, you install sudo apt-get install libopenblas-dev and set

       CONFIG_ARCH_BLAS=y

    • if you want to run GPU using ACL, see acl_driver.md for how to build ACL and set

       CONFIG_ACL_GPU=y
 ACL_ROOT = /home/firefly/ComputeLibrary(your ACL root)

    • Serializer support: by default, caffe model serializer option is valid CONFIG_CAFFE_SUPPORT=y.

      If you want to support mxnet, tensorflow, onnx serializer, you uncomment the options

       #CONFIG_MXNET_SERIALIZER=y
 #CONFIG_ONNX_SERAILIZER=y
 # CONFIG_TF_SERIALIZER=y

2. Build

cd ~/tengine
make

3. Run Demo

Tengine also provides some example programs for tests, and you can easily validate whether your Tengine is successfully built by running these test programs and inspecting the results.

3.1 Run SqueezeNet

	./build/tests/bin/bench_sqz -r1

    `-r1` means repeat one time.

Output message:

    0.2763 - "n02123045 tabby, tabby cat"
    0.2673 - "n02123159 tiger cat"
    0.1766 - "n02119789 kit fox, Vulpes macrotis"
	0.0827 - "n02124075 Egyptian cat"
    0.0777 - "n02085620 Chihuahua"

3.2 Run MobileNet

    ./build/tests/bin/bench_mobilenet -r1

Output message:

	8.5976 - "n02123159 tiger cat"
    7.9550 - "n02119022 red fox, Vulpes vulpes"
	7.8679 - "n02119789 kit fox, Vulpes macrotis"
    7.4274 - "n02113023 Pembroke, Pembroke Welsh corgi"
    6.3647 - "n02123045 tabby, tabby cat"

For more information about the performance test of Tengine, please refer to the documentation of benchmark.

Please visit exmaples for applications on classification/detection etc.