Dimensional Data

Large language models (LLMs) are revolutionizing the way we interact with software by combining deep learning techniques with powerful computational resources. While this technology is exciting, many are also concerned about how LLMs can generate false, outdated, or problematic information, and how they sometimes even hallucinate (generating information that doesn’t exist) so convincingly. Thankfully, we can immediately put one rumor to rest. According to Alireza Goudarzi, senior researcher of machine learning (ML) for GitHub Copilot: “LLMs are not trained to reason. They’re not trying to understand science, literature, code, or anything else. They’re simply trained to predict the next token in the text.” Let’s dive into how LLMs come to do the unexpected, and why. This blog post will provide comprehensive insights into LLMs, including their training methods and ethical considerations. Our goal is to help you gain a better understanding of LLM capabilities and how they’ve learned to master language, seemingly, without reasoning. What are large language models? LLMs are AI systems that are trained on massive amounts of text data, allowing them to generate human-like responses and understand natural language in a way that traditional ML models can’t. “These models use advanced techniques from the field of deep learning, which involves training deep neural networks with many layers to learn complex patterns and relationships,” explains John Berryman, a senior researcher of ML on the GitHub Copilot team. What sets LLMs apart is their proficiency at generalizing and understanding context. They’re not limited to pre-defined rules or patterns, but instead learn from large amounts of data to develop their own understanding of language. This allows them to generate coherent and contextually appropriate responses to a wide range of prompts and queries. And while LLMs can be incredibly powerful and flexible tools because of this, the ML methods used to train them, and the quality—or limitations—of their training data, can also lead to occasional lapses in generating accurate, useful, and trustworthy information. Deep learning The advent of modern ML practices, such as deep learning, has been a game-changer when it comes to unlocking the potential of LLMs. Unlike the earliest language models that relied on predefined rules and patterns, deep learning allows these models to create natural language outputs in a more human-like way. “The entire discipline of deep learning and neural networks—which underlies all of this—is ‘how simple can we make the rule and get as close to the behavior of a human brain as possible?’” says Goudarzi. By using neural networks with many layers, deep learning enables LLMs to analyze and learn complex patterns and relationships in language data. This means that these models can generate coherent and contextually appropriate responses, even in the face of complex sentence structures, idiomatic expressions, and subtle nuances in language. While the initial pre-training equips LLMs with a broad language understanding, fine-tuning is where they become versatile and adaptable. “When developers want these models to perform specific tasks, they provide task descriptions and examples (few-shot learning) or task descriptions alone (zero-shot learning). The model then fine-tunes its pre-trained weights based on this information,” says Goudarzi. This process helps it adapt to the specific task while retaining the knowledge it gained from its extensive pre-training. But even with deep learning’s multiple layers and attention mechanisms enabling LLMs to generate human-like text, it can […]

vcpkg 2023.10.19 Release: Export for Manifests, Documentation Improvements, and More… Augustin Popa November 3rd, 20230 0 The 2023.10.19 release of the vcpkg package manager is available. This blog post summarizes changes from August 10th, 2023 to October 19th, 2023 for the Microsoft/vcpkg, Microsoft/vcpkg-tool, and Microsoft/vcpkg-docs GitHub repos. Some stats for this period: 53 new ports were added to the open-source registry. If you are unfamiliar with the term ‘port’, they are packages that are built from source and are typically C/C++ libraries. 729 updates were made to existing ports. As always, we validate each change to a port by building all other ports that depend on or are depended by the library that is being updated for our nine main triplets. There are now 2,318 total libraries available in the vcpkg public registry. 34 contributors submitted PRs, issues, or participated in discussions in the main repo. The main vcpkg repo has over 5,700 forks and 19,900 stars on GitHub.   Key changes Notable changes for this release are summarized below. vcpkg export now supports manifest mode The vcpkg export command can be used to export built packages from the installed directory to a standalone SDK. A variety of formats are supported, including NuGet, a zip, or a raw directory. The SDK contains all prebuilt binaries for the selected packages, their transitive dependencies, and integration files such as CMake toolchain or MSBuild props/targets. This command is useful for developers who want to export their dependencies to a portable format for their end users to consume, when those end users do not have vcpkg. Now, this command is supported for manifest-based (vcpkg.json) projects. Summary of changes: vcpkg export in manifest mode exports everything in the vcpkg_installed directory In manifest mode, the export command emits an error and fails when port:triplet arguments are provided, as they are not allowed in manifest mode. Added a guard to exit with an error message when the installed directory is empty. Previously, it just failed silently. Made –output-dir mandatory in manifest mode. Documentation for vcpkg export in manifest mode PR: Microsoft/vcpkg-tool#1136   Implemented default triplet changes announced earlier this year In a previous blog post, we announced that we would be changing the default behavior for commands that accept a triplet as an option but are not provided one. This change is now live. The default triplet assumed is no longer always x86-windows but will instead use a triplet inferred from your CPU architecture and operating system. PR: Microsoft/vcpkg-tool#1180   Improvements to vcpkg help The documentation provided when running vcpkg help or vcpkg help has been updated. This should make it easier to explore vcpkg without having to go to the documentation on Microsoft Learn and also improves the autocompletion (Tab) experience for the tool. Below are some screenshots of the new experience. Before (left) and after (right) when running vcpkg help (or not providing any commands to vcpkg): Summary of changes: We now show a more complete list of available commands and options. Available commands now organized by category and sorted alphabetically. Added a link to our online vcpkg documentation. Cleaned up some of the wording.   Before (left) and after (right) when running vcpkg help install: Summary of changes: Added synopsis describing what the command does. In some cases, added additional examples for different usage […]