1 paper across 1 session
We demonstrate for the first time, fully quantized training of a 7B LLM using FP4 format.
