ilnmtlbnm

Amazing!

I encountered the same type of death freeze when trying (and failing) to run models in browser tabs, but didn't spend much time trying to understand how severe it is.

Hope they don't disable WebGPU...

"reasoning models are dynamical systems with fractal basins"

I've got a cli version that's about 60MB total, no dependencies, including the (nano, for now, will add tiny soon) model -> https://github.com/idle-intelligence/tts-web

Not sure I understand what you're asking: there are faster than realtime models that you can use, locally or as APIs.

I've been playing with running small models in browser tabs for some time, and finally decided to open some of it.

Added kitten (nano only, for now, will move on to mini) to my "web tts thing": https://github.com/idle-intelligence/tts-web

demo: https://idle-intelligence.github.io/tts-web/web/

low latency speech-to-text -> LLM -> text-to-speech

From the readme:

"This project is a reimplementation of the nice MJML markup language in Rust."

Take a look at : https://www.softwareheritage.org/ They've made it their mission to collect and preserve source code.

This is part 2 of a post on "Building a machine reading comprehension system using the latest advances in deep learning for NLP".

It uses HugginFace's BERT PyTorch implementation to do question answering.

Part 1 was submitted here : https://news.ycombinator.com/item?id=20820016