注册并分享邀请链接,可获得视频播放与邀请奖励。

Ant Ling (@AntLingAGI) “We recently released a paper showing that UFP4, our uniform-grid FP4 training re” — TopicDigg

Ant Ling 的个人资料封面
Ant Ling 的头像
Ant Ling
@AntLingAGI
加入 April 2025
0 正在关注    0 粉丝
We recently released a paper showing that UFP4, our uniform-grid FP4 training recipe, stays closer to BF16 than strong E2M1 baselines across Dense 1.5B, MoE 7.9B, and MoE 124B long-run pretraining. The key insight: FP4 training quality is not only about bit width, but also grid geometry.
显示更多
0
3
114
8
转发到社区