Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention

Add code
Jul 05, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: