AI Deep Learning

AI Deep Learning

Up to 25 Times Faster

Install a comprehensive Python and GPU powered deep learning suite. The Nvidia RTX tensor cores can accelerate deep learning tasks by up to 25x (2,500%) versus other popular developer laptops.

Please read the disclaimer before proceeding. This solution is updated regularly. Please write authorship with suggestions or requests.

Check Libraries

The Focus ships with CUDA, TensorRT, cuDNN, OpenGL, and Vulkan libraries pre-installed. This enables sophisticated compute and gaming out of the box. However, adding repositories or packages can alter and break the libraries. You may check and reset them as shown below:

/usr/lib/kfocus-001/bin/fixup_run.sh 0 1 1 # If libraries do not match, run the following: /usr/lib/kfocus-001/bin/fixup_run.sh

Run kfocus-001-conda

Use kfocus-001-conda to install an Anaconda suite. This is professionally vetted, comprehensive, and GPU accelerated. Open a terminal and enter the following:

kfocus-001-conda

Follow the prompts. You will be asked to install the following environments:

The script will not overwrite or delete data unless you tell it to do so. You may run it again to add environments as needed. For example, you may install only TensorFlow at first and then add PyTorch later. Run the AI suite at the end to compare to the results shown in the Benchmarks section.

When you wish to build on these Conda environments, we recommend using Pycharm. You may install this with the JetBrains Toolbox installed on all Kubuntu Focus laptops.

Using Nvidia On-Demand

When running Deep Learning workloads, you may notice that the software reports less available RAM than the GPU actually has. Part of this is because of driver overhead, but also because the GPU uses some memory for your display. One can instead use the Intel GPU for the display and use the Nvidia GPU for only Deep Learning or other computing tasks. To do this, start the Nvidia Settings app and then select Prime Profiles > Nvidia On-Demand and reboot. This mode does not significantly improve battery life, so we recommend using Nvidia (Performance mode) or Intel (Power Saving Mode) as it suits your needs at other times.

Benchmarks

You may run comprehensive TensorFlow benchmarks at the end of the kfocus-001-conda script. Please compare to the official results. All tests are current as of November 2020.

RTX 2080 Super[1]RTX 2070[1]RTX 2060[1]Titan X[2]Titan XP W10[2]Titan XP Ubuntu[2]
Inference9,6268,8817,7857,4439,74111,948
Training9,7679,2958,6717,90810,37512,922
AI Score19,39318,17616,87815,35120,08924,870

[1] Kubuntu Focus 2.2Kg, 20mm thick

[2] Unconstrained Desktop GPU

Troubleshooting

Q: Starting Nvidia Visual Profiler results in an error. How can I fix this?

A: One must install Java 8 and then call nvvp with the correct path.

sudo install openjdk-8-jdk nvvp -vm /usr/lib/jvm/java-8-openjdk-amd64/jre/bin/java

Revisions

This is a partial revision history. See the git repository for all entries.

Disclaimer

We try hard to provide a useful solution validated by professionals. However, we cannot anticipate every situation, and therefore cannot guarantee this procedure will work for your needs. Always backup your data and test the solution to determine the correct procedure for you.

THIS SOLUTION IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOLUTION, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

KFocus Logo