GPU

Compiling ffmpeg with NVIDIA GPU Hardware Acceleration on Ubuntu 20.04LTS

 in Applications  tagged cuda / ffmpeg / GPU / nvidia / ubuntu by

Please note that the following commands were executed on a system that already had CUDA support so we might be missing a step or two to enable NVIDIA CUDA support.

Install necessary packages

sudo apt-get install build-essential yasm cmake libtool libc6 libc6-dev unzip wget libnuma1 libnuma-dev nvidia-cuda-toolkit;

Clone and install ffnvcodec

git clone https://git.videolan.org/git/ffmpeg/nv-codec-headers.git;
cd nv-codec-headers;
sudo make install;
cd -;

Clone and compile FFmpeg’s public GIT repository with NVIDIA GPU hardware acceleration

git clone https://git.ffmpeg.org/ffmpeg.git ffmpeg/;
cd ffmpeg;
./configure --enable-nonfree --enable-cuda-nvcc --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64 --disable-static --enable-shared;
make -j 8;
sudo make install;

SUCCESS!

After performing the above steps, we were able to process media using ffmpeg without stressing our CPU! The workload was transferred to the GPU!

How to check if PyTorch is using the GPU?

 in Python  tagged GPU / python / pytorch by

The following basic code, will import PyTorch into a project and test if GPU capabilities are enabled and available.

import torch

# Should produce "True"
torch.cuda.is_available()

# Should produce the number of available devices, if you have one device it should produce the value "1"
torch.cuda.device_count()

# If there is a device and it is the first one, it should produce the "0"
torch.cuda.current_device()

# Assuming that there is at least one device, with the following two commands we will get some information on the first available device

# Should produce something similar to "<torch.cuda.device object at 0x7f12b1a298d0>"
torch.cuda.device(0)

# Should produce something similar to "'GeForce GTX 1050 Ti'"
torch.cuda.get_device_name(0)

We are using Ubuntu 20.04 LTS and the NVidia drivers were installed automatically during installation.

Quick note on setting up our programming environment for Coursera.org “DeepLearning.AI TensorFlow Developer Professional Certificate” on Ubuntu 18.04LTS

 in Programming  tagged DeepLearning.AI / GPU / jupyter / LD_LIBRARY_PATH / libcublas / nvidia / TensorFlow / ubuntu by

Recently we were working on some local Jupyter notebooks for the “DeepLearning.AI TensorFlow Developer Professional Certificate“.

To prepare for the setup with GPU, we followed the instructions at https://www.tensorflow.org/install/gpu. But as we were going through the notebooks we noticed the following error at the terminal:

W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'libcublas.so.10'; dlerror: libcublas.so.10: cannot open shared object file: No such file or directory

Even though libcublas.so.10 was installed it was not loading. Using find was saw that it was installed and available at the folder /usr/local/cuda-10.2/targets/x86_64-linux/lib/ but it was ignored.

To fix this, we added a new entry in ~/.profile as follows:

if [ -d "/usr/local/cuda-10.2/targets/x86_64-linux/lib/" ]; then
    export LD_LIBRARY_PATH=/usr/local/cuda-10.2/targets/x86_64-linux/lib/:${LD_LIBRARY_PATH}
fi

And then executed source ~/.profile from the terminal which we would start the jupyter-notebook command.

After this, library was loaded as expected.