Training and inference of large language models using 8-bit floating point

Open in new window