77 posts tagged “webassembly”
2023
The WebAssembly Go Playground (via) Jeff Lindsay has a full Go 1.21.1 compiler running entirely in the browser.
Perplexity: interactive LLM visualization (via) I linked to a video of Linus Lee’s GPT visualization tool the other day. Today he’s released a new version of it that people can actually play with: it runs entirely in a browser, powered by a 120MB version of the GPT-2 ONNX model loaded using the brilliant Transformers.js JavaScript library.
Wikipedia search-by-vibes through millions of pages offline (via) Really cool demo by Lee Butterman, who built embeddings of 2 million Wikipedia pages and figured out how to serve them directly to the browser, where they are used to implement “vibes based” similarity search returning results in 250ms. Lots of interesting details about how he pulled this off, using Arrow as the file format and ONNX to run the model in the browser.
WebLLM supports Llama 2 70B now. The WebLLM project from MLC uses WebGPU to run large language models entirely in the browser. They recently added support for Llama 2, including Llama 2 70B, the largest and most powerful model in that family.
To my astonishment, this worked! I used a M2 Mac with 64GB of RAM and Chrome Canary and it downloaded many GBs of data... but it worked, and spat out tokens at a slow but respectable rate of 3.25 tokens/second.
Building a Signal Analyzer with Modern Web Tech (via) Casey Primozic’s detailed write-up of his project to build a spectrogram and oscilloscope using cutting-edge modern web technology: Web Workers, Web Audio, SharedArrayBuffer, Atomics.waitAsync, OffscreenCanvas, WebAssembly SIMD and more. His conclusion: “Web developers now have all the tools they need to build native-or-better quality apps on the web.”
Web LLM runs the vicuna-7b Large Language Model entirely in your browser, and it’s very impressive
A month ago I asked Could you train a ChatGPT-beating model for $85,000 and run it in a browser?. $85,000 was a hypothetical training cost for LLaMA 7B plus Stanford Alpaca. “Run it in a browser” was based on the fact that Web Stable Diffusion runs a 1.9GB Stable Diffusion model in a browser, so maybe it’s not such a big leap to run a small Large Language Model there as well.
[... 2,276 words]AI photo sorter (via) Really interesting implementation of machine learning photo classification by Alexander Visheratin. This tool lets you select as many photos as you like from your own machine, then provides a web interface for classifying them into labels that you provide. It loads a 102MB quantized CLIP model and executes it in the browser using WebAssembly. Once classified, a “Generate script” button produces a copyable list of shell commands for moving your images into corresponding folders on your own machine. Your photos never get uploaded to a server—everything happens directly in your browser.
Could you train a ChatGPT-beating model for $85,000 and run it in a browser?
I think it’s now possible to train a large language model with similar functionality to GPT-3 for $85,000. And I think we might soon be able to run the resulting model entirely in the browser, and give it capabilities that leapfrog it ahead of ChatGPT.
[... 1,751 words]Web Stable Diffusion (via) I just ran the full Stable Diffusion image generation model entirely in my browser, and used it to generate an image (of two raccoons eating pie in the woods, see “via” link). I had to use Google Chrome Canary since this depends on WebGPU which still isn’t fully rolled out, but it worked perfectly.
Weeknotes: A bunch of things I learned this week, plus datasette-explain
The Datasette table view refactor, JSON redesign and ?_extra= continues this week, mainly in this ongoing pull request and this tracking issue.
PocketPy. PocketPy is “a lightweight(~5000 LOC) Python interpreter for game engines”. It’s implemented as a single C++ header which provides an impressive subset of the Python language: functions, dictionaries, lists, strings and basic classes too. There’s also a browser demo that loads a 766.66 KB pypocket.wasm file (240.72 KB compressed) and uses it to power a basic terminal interface. I tried and failed to get that pypocket.wasm file working from wasmer/wasmtime/wasm3—it should make a really neat lightweight language to run in a WebAssembly sandbox.
Python Sandbox in Web Assembly (via) Jim Kring responded to my questions on Mastodon about running Python in a WASM sandbox by building this repo, which demonstrates using wasmer-python to run a build of Python 3.6 compiled to WebAssembly, complete with protected access to a sandbox directory.
2022
talk.wasm (via) “Talk with an Artificial Intelligence in your browser”. Absolutely stunning demo which loads the Whisper speech recognition model (75MB) and a GPT-2 model (240MB) and executes them both in your browser via WebAssembly, then uses the Web Speech API to talk back to you. The result is a full speak-with-an-AI interface running entirely client-side. GPT-2 sadly mostly generates gibberish but the fact that this works at all is pretty astonishing.
Microsoft Flight Simulator: WebAssembly (via) This is such a smart application of WebAssembly: it can now be used to write extensions for Microsoft Flight Simulator, which means you can run code from untrusted sources safely in a sandbox. I’m really looking forward to more of this kind of usage—I love the idea of finally having a robust sandbox for running things like plugins.
PyScript Updates: Bytecode Alliance, Pyodide, and MicroPython. Absolutely huge news about Python on the Web tucked into this announcement: Anaconda have managed to get a version of MicroPython compiled to WebAssembly running in the browser. Pyodide weighs in at around 6.5MB compressed, but the MicroPython build is just 303KB—the size of a large image. This makes Python in the web browser applicable to so many more potential areas.
About the sqlite3 WASM/JS Subproject. SQLite now maintains an official WebAssembly build. It’s influenced by sql.js but is a fresh implementation with its own API design. It also supports Origin-Private FileSystem (OPFS)—a very new standard which doesn’t yet have wide browser support that allows websites to save and load files using a dedicated folder on the host machine.
libsql (via) A brand new Apache 2 licensed fork of SQLite. The README explains the rationale behind the project: SQLite itself is open source but not open contribution, and this fork aims to try out new ideas. The most interesting to me so far is a plan to support user defined functions implemented in WebAssembly. The project also intends to use Rust for new feature development.
mod_wasm: run WebAssembly with Apache (via) Brand new Apache module from a team at VMWare: mod_wasm builds on top of wasmtime to let you write WebAssembly programs that are exposed to the world by Apache, using a mechanism that looks similar to old CGI scripts (headers passed in environment variables, request body sent to standard input). Includes a demo Docker image that runs using Python-compiled-to-WebAssembly.
Fastly Compute@Edge JS Runtime (via) Fastly’s JavaScript runtime, designed to run at the edge of their CDN, uses the Mozilla SpiderMonkey JavaScript engine compiled to WebAssembly.
Wasmtime Reaches 1.0: Fast, Safe and Production Ready! The Bytecode Alliance are making some confident promises in this post about the performance and stability of their Wasmtime WebAssembly runtime. They also highlight some exciting use-cases for WebAssembly on the server, including safe 3rd party plugin execution and User Defined Functions running inside databases.
Crunchy Data: Learn Postgres at the Playground (via) Crunchy Data have a new PostgreSQL tutorial series, with a very cool twist: they have a build of PostgreSQL compiled to WebAssembly which runs in the browser, so each tutorial is accompanied by a working psql terminal that lets you try out the tutorial contents interactively. It even has support for PostGIS, though that particular tutorial has to load 80MB of assets in order to get it to work!
Plugin support for Datasette Lite
I’ve added a new feature to Datasette Lite, my distribution of Datasette that runs entirely in the browser using Python and SQLite compiled to WebAssembly. You can now install additional Datasette plugins by passing them in the URL.
[... 865 words]Introducing sqlite-lines—a SQLite extension for reading files line-by-line (via) Alex Garcia wrote a brilliant C module for SQLIte which adds functions (and a table-valued function) for efficiently reading newline-delimited text into SQLite. When combined with SQLite’s built-in JSON features this means you can read a huge newline-delimited JSON file into SQLite in a streaming fashion so it doesn’t exhaust memory for a large file. Alex also compiled the extension to WebAssembly, and his post here is an Observable notebook post that lets you exercise the code directly.
Joining CSV files in your browser using Datasette Lite
I added a new feature to Datasette Lite—my version of Datasette that runs entirely in your browser using WebAssembly (previously): you can now use it to load one or more CSV files by URL, and then run SQL queries against them—including joins across data from multiple files.
[... 546 words]The State of WebAssembly 2022. Colin Eberhardt talks through the results of the State of WebAssembly 2022 survey. Rust continues to dominate as the most popular language for working to WebAssembly, but Python has a notable increase of interest.
Weeknotes: Datasette Lite, nogil Python, HYTRADBOI
My big project this week was Datasette Lite, a new way to run Datasette directly in a browser, powered by WebAssembly and Pyodide. I also continued my research into running SQL queries in parallel, described last week. Plus I spoke at HYTRADBOI.
[... 1,434 words]Datasette Lite: a server-side Python web application running in a browser
Datasette Lite is a new way to run Datasette: entirely in a browser, taking advantage of the incredible Pyodide project which provides Python compiled to WebAssembly plus a whole suite of useful extras.
[... 4,800 words]sqlite-utils 3.26.1 (via) I released sqlite-utils 3.36.1 with one tiny but exciting feature: I fixed its one dependency that wasn’t published as a pure Python wheel, which means it can now be used with Pyodide—Python compiled to WebAssembly running in your browser!
PyScript demos (via) PyScript was announced at PyCon this morning. It’s a new open source project that provides Web Components built on top of Pyodide, allowing you to use Python directly within your HTML pages in a way that is executed using a WebAssembly copy of Python running in your browser. These demos really help illustrate what it can do—it’s a fascinating new piece of the Python web ecosystem.
Mac OS 8 emulated in WebAssembly (via) Absolutely incredible project by Mihai Parparita. This is a full, working copy of Mac OS 8 (from 1997) running in your browser via WebAssembly—and it’s fully loaded with games and applications too. I played with Photoshop 3.0 and Civilization and there’s so much more on there to explore too—I finally get to try out HyperCard!