Install the necessary packages for StarLight

Installation
Bugs and solutions

Installation

Create a conda envirment. You can use the Tsinghua source for the conda and pip to accelerate installation.
```
conda create -n starlight python=3.6
conda activate starlight
```

Install PyTorch and cuDNN.

conda install pytorch=1.6.0 torchvision=0.7.0 cudatoolkit=10.2
conda install --channel https://conda.anaconda.org/nvidia cudnn=8.0.0

Install TensorRT.

# Go to https://developer.nvidia.com/compute/machine-learning/tensorrt
# Download TensorRT-7.1.3.4.Ubuntu-18.04.x86_64-gnu.cuda-10.2.cudnn8.0.tar.gz 
tar -zxf TensorRT-7.1.3.4.Ubuntu-18.04.x86_64-gnu.cuda-10.2.cudnn8.0.tar.gz
cd TensorRT-7.1.3.4/ 
pip install python/tensorrt-7.1.3.4-cp36-none-linux_x86_64.whl
pip install uff/uff-0.6.9-py2.py3-none-any.whl
pip install graphsurgeon/graphsurgeon-0.4.5-py2.py3-none-any.whl
  
# Test if the TensorRT is installed successfully
python
import tensorrt 
# No error mean success, we have summarized two common bugs in the next section

Install PYQT5 and PyQtWebEngine:

pip install pyqt5==5.12
pip install PyQtWebEngine==5.12

Install other packages.

pip install easydict opencv-python flask flask_cors gevent imageio pynvml pyyaml psutil matplotlib pycocotools Cython thop schema prettytable
pip install onnx==1.11.0 pycuda==2019.1.1 tensorboard==2.9.1 tqdm
pip install opencv-python pdf2image 

Bugs and solutions

If libnvinfer.so.7 or libcudnn.so.8 is missing when you import the tensorrt, simply specify there direction in the ~/.bashrc:

# search their direction
find / -name libnvinfer.so.7
find / -name libcudnn.so.8
# for libnvinfer.so.7
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/env/TensorRT-7.1.3.4/lib
# for libcudnn.so.8
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/root/anaconda3/envs/starlight/lib

If ImportError: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.27’ not found, build GLIBC_2.27 manually:

# download glibc-2.27
wget http://ftp.gnu.org/gnu/glibc/glibc-2.27.tar.gz
tar -zxf glibc-2.27.tar.gz
cd glibc-2.27
mkdir build
cd build/
../configure --prefix=/opt/glibc-2.17 # <-- where you install glibc-2.27
# if error for gawk/bison, install them using: sudo apt-get install gawk/bison
make -j <number of CPU Cores>  # You can find your <number of CPU Cores> by using `nproc` command
make install
# patch your Python
patchelf --set-interpreter /opt/glibc-2.17/lib/ld-linux-x86-64.so.2 --set-rpath /opt/glibc-2.17/lib/ /root/anaconda3/envs/starlight/bin/python

Install the necessary packages for StarLight

Table of contents

Installation

Bugs and solutions