Report

이승우, 이경준, 이중원

Report - Improve the accuracy for FP8

손동휘, 최종현

Report - Int8 GEMM CUTLASS implementation

문형준, 최수환

Report - Support Torch compile @ispobock

문킴, 서승현, 양준영

Report - TP+DP Attention + MOE

박주언, 박천명

Report - Support overlap scheduler with DP attention @merrymercy

마이클, 김형민

Report - Support nextn speculative decoding @ispobock

H100 server access

~/.ssh/config

Host h1001
  HostName h1001
  User USER_NAME

/etc/hosts

65.108.33.92 h1001