canzhiye/install.md

## install.md

      
    Raw
  

              install.md
            
          
    installing densepose

The installation guide on https://github.com/facebookresearch/DensePose seemed pretty lacking and I spent a bunch of time trying to figure out how to install so I kept track of my steps. Voila.

Go to your EC2 Dashboard
Launch a Ubuntu 16.04 instance
Choose an instance type that has a GPU. I think g2.2xlarge is the cheapest one.
Before you launch it, make sure you change the storage settings so that the root volume is 100GB not 8GB.
Launch it with an existing key pair if you have one, or create a new one if you don’t
SSH into you instance after it launches:
ssh -i ~/.ssh/YOUR_KEY_PAIR.pem ubuntu@YOUR-EC2-ADDRESS.compute.amazonaws.com


Once you are SSHed in, install a bunch of dependencies.
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
      build-essential \
      cmake \
      git \
      libgoogle-glog-dev \
      libgtest-dev \
      libiomp-dev \
      libleveldb-dev \
      liblmdb-dev \
      libopencv-dev \
      libopenmpi-dev \
      libsnappy-dev \
      libprotobuf-dev \
      openmpi-bin \
      openmpi-doc \
      protobuf-compiler \
      python-dev \
      python-pip  
      
sudo apt-get install python-setuptools
sudo pip install future numpy protobuf enum networkx cython

sudo apt-get install -y --no-install-recommends libgflags-dev


Now install CUDA
sudo apt-get update && sudo apt-get install wget -y --no-install-recommends
wget "http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb"
sudo dpkg -i cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda


Download cuDNN v6.0 for CUDA 8.0 here. You will have to make an NVIDIA developer account. Find the link that says "cuDNN v6.0 Library for Linux" and download it to your Downloads folder.
Open a new Terminal tab and copy it over to your EC2 instance.
scp -i ~/.ssh/YOUR_KEY_PAIR.pem ~/Downloads/cudnn-8.0-linux-x64-v6.0.tgz ubuntu@YOUR-EC2-ADDRESS.compute.amazonaws.com:~/

Go back to your EC2 instance tab and install.
sudo tar -xzf cudnn-8.0-linux-x64-v6.0.tgz -C /usr/local
rm cudnn-8.0-linux-x64-v6.0.tgz && sudo ldconfig


(skip this section if you remembered to edit your storage settings when launching the instance)
densepose is huge and the disk that comes with the g2.2xlarge is too small so you have to expand the disk size.
Open your EC2 Dashboard. Click on the instance you just launched. Find Block Devices, and click on "/dev/sda1" and then click on the EBS ID link. The click Actions, and select Modify Volume. I added 100GB -- not exactly sure how much you need. This takes a few minutes to take effect. Type lsblk in your Terminal to see if the disk has been expanded.
Next you need to expand your partition. Run sudo growpart /dev/xvda 1
I had to reboot the instance for this to take effect.

Run the following to install Caffe2 first
export PYTHONPATH=/usr/local
export PYTHONPATH=$PYTHONPATH:/home/ubuntu/pytorch/build
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

git clone --recursive https://github.com/pytorch/pytorch.git && cd pytorch
git submodule update --init

# Create a directory to put Caffe2's build files in
mkdir build && cd build

# Configure Caffe2's build
# This looks for packages on your machine and figures out which functionality
# to include in the Caffe2 installation. The output of this command is very
# useful in debugging.
cmake ..

sudo make install -j8

Run python pytorch/caffe2/python/operator_test/reshape_ops_test.py and make sure this works.

Install COCO dataset
cd
sudo git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI/
sudo make install

Install detectron
cd
git clone https://github.com/facebookresearch/detectron
sudo pip install -r detectron/requirements.txt 
cd detectron
make

Test to make sure detectron installed properly
cd
python detectron/detectron/tests/test_spatial_narrow_as_op.py 

Install densepose
git clone https://github.com/facebookresearch/densepose
sudo pip install -r densepose/requirements.txt
cd densepose && make
make ops

Test to make sure densepose works
python2 $DENSEPOSE/detectron/tests/test_zero_even_op.py


Enjoy :)
Fun thing I made with this