kingoflolz / swarm-jax Goto Github PK
View Code? Open in Web Editor NEWSwarm training framework using Haiku + JAX + Ray for layer parallel transformer language models on unreliable, heterogeneous nodes
Swarm training framework using Haiku + JAX + Ray for layer parallel transformer language models on unreliable, heterogeneous nodes
Hiya! I was talking with Skye over at google/jax#2108 (comment)
one issue with the code snippet you have at the top for transferring grads from TPU to CPU memory: device_put is a no-op inside jit (I think there may an issue about making this less of a gotcha, but I can't find it now). Using device_put with a CPU device is the right idea though, it just has to happen outside of a compiled function.
The code snippet Skye is referring to was copied from this codebase. So I thought I should open an issue, because it sounds like that code is nonfunctional.
I’m posting this from my phone, but I’ll add more details in a little while.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.