Will AI development become more accessible? What will change with the release of the Arm version of PyTorch?
Hello everyone! The world of AI is evolving rapidly, and new news comes in every day. This time, we will explain the "Arm version of PyTorch," which is one of the most noteworthy topics among the latest AI information, in an easy-to-understand manner even for AI beginners!
Copilot+ PC and AI Evolution
Did you know that Microsoft has announced a new PC called Copilot+ PC? This is a PC for the AI era, equipped with features to make AI (artificial intelligence) processing more efficient on the PC. It is designed to allow developers to make the most of the PC's AI capabilities using a tool called Copilot Runtime.
Until now, in order to run a large-scale AI model, it was necessary to use the cloud (a huge computer resource on the Internet). However, with Copilot+ PC, you can run a relatively small AI model on your PC, so there is less concern about personal information leaking to the outside. It's like doing AI research in your own home!
What is an NPU? The secret weapon that accelerates AI
The "NPU (Neural Processing Unit)" plays a key role here. This is a dedicated processor (the core part of a computer) specialized for AI calculation processing. It could be said to be the third processor alongside the CPU (Central Processing Unit) and GPU (Graphics Processing Unit). The NPU has the power to greatly improve AI performance, especially since it can perform the calculations required for deep learning at high speed. To put it in perspective, it's like having a special calculator dedicated to AI research built into your computer!
Arm version of PyTorch is here! What's so great about it?
And this time, the noteworthy news is the arrival of the "Arm version of PyTorch"! PyTorch is one of the very popular frameworks (a set of useful tools that help development) for developing AI models. It is especially popular among researchers. This Arm version of PyTorch is optimized for the Arm processor (a low-energy CPU often used in smartphones and tablets) installed in the Copilot+ PC.
Until now, AI development has typically involved using a PC equipped with a high-performance GPU (graphics board). However, with the arrival of the Arm version of PyTorch, it has become easy to develop and experiment with AI models even on a PC equipped with an Arm processor. This is a big step forward in broadening the scope of AI development.
What can you do with PyTorch?
PyTorch makes it easy to handle tensors (multidimensional arrays, or simply put, data with many numbers) and build neural networks (AI models that mimic the neural circuits of the human brain). For example, if you want to create an image recognition AI, you can use PyTorch to read image data as tensors and train it using a neural network.
PyTorch is also widely used in Hugging Face, an AI model sharing platform. Hugging Face has published a variety of AI models developed by researchers around the world, and with PyTorch you can easily try out these models and retrain them with your own data.
How to install PyTorch for Arm
To install the Arm version of PyTorch, you need to do some preparation. First, you need to install development tools such as Visual Studio Build Tools (including the C++ compiler) and Python (programming language). After that, use a tool called pip to install PyTorch. For detailed instructions, please refer to the official Microsoft blog.
Let's actually run the AI model!
Once you have installed the Arm version of PyTorch, let's try running an AI model right away! Microsoft has released sample code, so we recommend using that. The sample code allows you to download an image generation AI model called Stable Diffusion and build a simple web server and UI (user interface).
Using this UI, you can enter prompts (instructions) for the image you want to generate and adjust parameters (settings). When you actually try to generate an image, you'll see that even the Arm processor is capable of fairly complex processing.
Future prospects
At present, the Arm version of PyTorch cannot fully utilize the NPU (a processor dedicated to AI processing). However, in the future, it is expected that support for NPUs will improve, enabling even faster AI processing. If that happens, Copilot+ PC will become an even more attractive platform for AI developers.
Summary and Thoughts
The arrival of the Arm version of PyTorch is news that will greatly expand the possibilities of AI development. Until now, AI development required an expensive PC equipped with a high-performance GPU, but by making it easier and more accessible, more people will be able to enter the world of AI. I tried it myself, and was impressed by how easily I could run an AI model, much easier than I had imagined!
This article is based on the following original articles and is summarized from the author's perspective:
Running PyTorch on an Arm Copilot+ PC