Skip to main content

Posts

2025

A minor point about precision
·1205 words
Things I learned digging into 5090 perf
·1040 words
Various approaches to parallelizing Muon
·1960 words

2024

Testing the 4090 48GB
·1050 words
Visualizing 6D Mesh Parallelism
·3341 words
Why reduction precision matters
·1054 words