NVIDIA recently released the 12.0 version of the CUDA Toolkit. This release, which focused on new programming models and CUDA application acceleration through new hardware capabilities, was the first significant update in a long time. After this update, we can now target CUDA custom code, improved libraries, and developer tools that provide architecture-specific features and instructions in the NVIDIA Hopper and NVIDIA Ada Lovelace architectures. NVIDIA's parallel computing platform, CUDA (Compute Unified Device Architecture), was created for general computing and is the main basis for GPGPU. It is a layer of software that gives compute kernels direct access to the virtual instruction set of GPUs as well as parallel computational components.

This article installs the drivers and programs that are needed to use NVIDIA GPUs to train models and run batch inferences. It downloads, unzips, and moves the CuDNN and TensorRT files into the CUDA directory. It also configures, builds, and runs the BlackScholes sample to test the GPU. This section joins the NVIDIA Developer Program and downloads the CuDNN library and unzips and moves the files into the CUDA directory. This section downloads the TensorRT library and unzips and moves the files into the CUDA directory and installs several required python programs. This section configures, builds, and runs the BlackScholes sample.

It is always convoluted and challenging to install a CUDA toolkit and library that needs to interact with your NVIDIA GPU on an Ubuntu machine. However, if done right, the CUDA toolkit harnessing your NVIDIA GPU can be a great tool that can harness the power of GPU to produce fast applications. The basic requirement for following instructions in this article is a computer with Ubuntu 20.04 installed with an NVIDIA GPU. In my case, it was NVIDIA GeForce GTX 1650 Ti. Further, at the time of writing this article, I installed the latest version of the CUDA toolkit which was CUDA Toolkit 11.3.

Visual studio is required for the installation of Nvidia CUDA Toolkit (this prerequisite is referred to here). If you attempt to download and install CUDA Toolkit for Windows without having first installed Visual Studio, you get the message shown in Figure 1. At the time of writing, the most recent version of Visual Studio (which is free) is the Visual Studio Express Community Version 2017, shown in Fig 2. You can get previous versions of Visual Studio for free by joining "Visual Studio Dev Essentials" and then searching for the version of Visual Studio you want. Once you have downloaded Visual Studio Express, its installation is straightforward.

Lately, I have been consolidating my experiences of working in different ML projects. I will tell this story from the lens of my recent NLP project to classify phrases into categories -- A multiclass single label problem. Making AI teams is quite tricky. If you don't have the skillsets inside your company, you have to plan hiring. Since every project has a start and end time, it's difficult to have the entire team from the start.

Lately, I have been consolidating my experiences of working in different ML projects. I will tell this story from the lens of my recent NLP project to classify phrases into categories -- A multiclass single label problem. Making AI teams is quite tricky. If you don't have the skillsets inside your company, you have to plan hiring. Since every project has a start and end time, it's difficult to have the entire team from the start.

In recent years, there has been significant progress in the field of machine learning. Much of this progress can be attributed to the increasing use of graphics processing units (GPUs) to accelerate the training of machine learning models. In particular, the extra computational power has lead to the popularization of deep learning -- the use of complex, multi-level neural networks to create models, capable of feature detection from large amounts of unlabeled training data. GPUs are great for deep learning because the type of calculations they were designed to process are the same as those encountered in deep learning. Images, videos, and other graphics are represented as matrices so that when you perform any operation, such as a zoom-in effect or a camera rotation, all you are doing is applying some mathematical transformation to a matrix.

I like to watch long-term technology and business trends and watch as they shape the products and services that I get to use and to write about. As I was preparing to write today's post, three such trends came to mind: As the industry pushes forward in accord with these trends, a couple of interesting challenges have surfaced over the past decade or so. Again, here's a quick list (yes, I do think in bullet points): The GPU (Graphics Processing Unit) was born of these trends, and addresses many of the challenges! Processors have reached the upper bound on clock rates, but Moore's Law gives designers more and more transistors to work with. Those transistors can be used to add more cache and more memory to a traditional architecture, but the von Neumann Bottleneck limits the value of doing so. On the other hand, we now have large markets for specialized hardware (gaming comes to mind as one of the early drivers for GPU consumption).

The purpose of this blog post is to demonstrate how to install the Keras library for deep learning. Let me start by saying that Keras is my favorite deep learning Python library. It's a minimalist, modular neural network library that can use either Theano or TensorFlow as a backend. Furthermore, the primary motivation behind Keras really resonates with me: you should be able to experiment super quickly -- going from idea to result, as fast as possible. Coming from a world that mixes both academia and entrepreneurship, the ability to iterate quickly is extremely valuable, especially in the deep learning world where it can take days to weeks to train just a single model.