Running modern LLMs in the browser using Three.js and WebGPU (ben3d.ca via hn)
model roundup
GPT 2
-
Running LLMs in the Browser with Three.js A tour of how language models generate text, how GPT-2 and newer Phi and Qwen architectures differ, and how Three-LLM implements their inference graphs with Three.js TSL compute shaders. Ben Housto…
-
What Makes LLM Tokenization Slow? (healeycodes.com via hn)
LLMs receive sequences of token IDs instead of raw text. For example, the GPT-2 tokenizer encodes what's the weather in goldshire?
-
Show HN: Heaviside-1, a foundation model for electromagnetism (fields-studio.arenaphysica.com via hn)
Heaviside-1 was trained on 500B field samples from over 250K unique physical layout designs, and is approximately the scale of GPT-2. It predicts a full design's fields ~10^5x faster than a commercial full-wave solver, with S-parameter mag…