Guardado en:
Detalles Bibliográficos
Autores principales: Mishra, Asit, Stosic, Dusan, Layton, Simon, Micikevicius, Paulius
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2506.08027
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Tabla de Contenidos:
  • Using fewer bits to represent model parameters and related tensors during pre-training has become a required technique for improving GPU efficiency without sacrificing accuracy. Microscaling (MX) formats introduced in NVIDIA Blackwell generation of GPUs represent a major advancement of this technique, making it practical to combine narrow floating-point data types with finer granularity per-block scaling factors. In turn, this enables both quantization of more tensors than previous approaches and more efficient execution of operations on those tensors. Effective use of MX-formats requires careful choices of various parameters. In this paper we review these choices and show how MXFP8-E4M3 datatype and a specific number conversion algorithm result in training sessions that match those carried out in BF16. We present results using models with up to 8B parameters, trained on high-quality datasets of up to 15T tokens.