Project Website --- www.TECHNOPREMIUM.com
Buy DLBT SSD ready -- BUY HERE
Check our Deep learning blog --- DLBT_BLOG
DeepLearning Benchmark Tool is an application whose purpose is measuring the performance of a particular hardware in the specific task of running a deep learning model. By running, it implies the device should do both the training of the model, and the prediction while the parameters are being adjusted.
One of the main areas of application this software has is serving as a test bench for Graphical Processing Units (GPU) in the task of Machine Learning. This is mainly because, as has been shown in last few years, these devices have a parallelism quality - GPUs can have more than 2000 cores- that makes them much better than Central Processing Units (CPU) at running Deep Learning Models. This application, however, lets the user choose whether they would prefer to run the test on a CPU or a GPU.
- Step 1
Deep Learning specific hardware needs to have a great parallelizing quality, this means it should be able to do multiples calcutations at the same time. Graphical Processing Units from NVidia have all the support needed to be used in your Artificial Intelligence Projects, we help you by saying exactly how good they are for Deep Learning
- Step 2
We use a Convoulutional Neural Network model that is training from an image dataset. The model should learn the features from each class in the images. In our app you decide how far into the training you want to go.
- step 3
We take several metrics into account when deciding just how good is your device for Deep Learning. For example, we see how efficient is it predicting the result, and how fast it can learn the features from the training datasets. The better the hardware the higher the scores you get, and with the Community results built into the app, now you can show to the world how powerful is your hardware and see others so you can understand what is the best hardware to buy for your next DL workstation machine.
- We know how frustrating it is to install all dependencies and frameworks to run a simple benchmark; most of Github benchmarks required basic knowledge of the command line and docker commands. DLBT makes things different from a User-friendly interface; everyone can now run Deep Learning Benchmarks to see the performance on their Hardware, Stress test at indeed maximum potential, and understand what Hardware is better than the other to make the right purchase for your use case.
- Test not with just one but multiple GPU support without writing a line of code or installing any libraries, everything is ready to go, select the amount of GPUs you want to use and click RUN
- We support all the most popular models, so you can check your Hardware and make sure what is the best configuration for you. ٠ Resnet50 ٠ Resnet152 ٠ Inception v3٠ Inception v4 ٠ VGG16
- Usually, you need to know what parameters affect a model when training, let’s say batch size, iterations per second, the model you want to use, using XLA or not, FP16 or FP32. Well, all this is now integrated into our UI, and with a click, you can select all these settings without memorizing them and understand the command line.
- DLBT is not just a benchmark for Deep learning/ Machine learning users, any other user can run the benchmark as well, since most benchmark ( Gaming ) use a part of the GPU you never know how reliable the GPU is, with DLBT you will push the Hardware to the limit. It will utilize all the Vram, TDP, Graphical cores, CUDA cores, and even the new Tensor cores at 100%, so after 1 hour testing or even more with stress-test, you will understand if your Hardware it's ready to do their job.
- We have built-in into our app an easy way to share your score online, show to the world your specs, and be the first in the list. This helps others to understand what GPU is better for a particular task without having the GPUs so that they can make the right decision.
- Our DLBT app runs entirely on GUI mode for you just with clicks to be able to run the most popular Deep Learning models.
- Currently, we have full support for Ubuntu 18.04, but if you have Nvidia driver installed in your Linux distro, our app will work with no problem since, yes, it's built-in with DOCKER, making it even more straightforward for deployment.
- We are a group of passionate Developers, and our mission is to bring Deep learning to everyone, making these tools accessible to the ordinary public, make it easier and fun to try Deep learning. We also collaborate with Businesses to integrate our Devops and DL expertise into their business models.
- Linux distro - since its a docker image will work on any Linux distro, but its currently tested on Ubuntu 18.04 and CentOS.
- Nvidia driver 418+
- Docker installed with nvidia-runtime
If you don't have this requirement, please contact us for further instructions EMAIL or buy a ready to go DLBT SSD that you can use in multiple systems without messing around with yours ---- BUY HERE
Clone this repo to your local machine.
git clone https://github.com/technopremium/dlbt_public.git
cd dlbt_public/
sudo chmod +x start.sh
sudo chmod +x launcher
./start.sh
./launcherThis will open the DLBT APP -- See picture below
Video Guides available HERE
email - support@technopremium.com Website - www.TECHNOPREMIUM.com
Facebook Instagram Youtube Linkedin Twitter Pinterest Reddit Medium