I drew heavy inspiration from this video “Not Knot” by the Geometry Center / Geometry Supercomputer Project in 1991, directed by Charlie Gunn and Delle Maxwell.
[tracks, filter]
。业内人士推荐Line官方版本下载作为进阶阅读
FT Digital Edition。关于这个话题，PDF资料提供了深入分析
Rank-3 factorization, shared-A tied-KV, RMSNorm, grokking