DeepSpeed
[zero] revert PR #3166, it disabled grad clip for bf16
#3790
Merged
Overview
Commits
29
Changes
View On GitHub
Commits
  • zero++ tutorial PR (#3783)
    HeyangQin committed 2 years ago
  • [Fix] _conv_flops_compute when padding is a str and stride=1 (#3169)
    zhiruiluo committed 2 years ago
  • fix interpolate flops compute (#3782)
    cli99 committed 2 years ago
  • use `Flops Profiler` to test `model.generate()` (#2515)
    CaffreyR committed 2 years ago
  • revert PR #3166, it disabled grad clip for bf16
    jeffra committed 2 years ago
  • ensure no loss scaling for non-fp16 dtypes
    jeffra committed 2 years ago
  • revert PR #3611 (#3786)
    jeffra committed 2 years ago
  • bump to 0.9.6
    jeffra committed 2 years ago
  • Merge branch 'master' into revert-3166
    jeffra committed 2 years ago
  • ZeRO++ chinese blog (#3793)
    HeyangQin committed 2 years ago
  • remove staging trigger (#3792)
    jeffra committed 2 years ago
  • DeepSpeed-Triton for Inference (#3748)
    stephen-youn committed 2 years ago
  • ZeRO++ (#3784)
    HeyangQin committed 2 years ago
  • adding zero++ to navigation panel of deepspeed.ai (#3796)
    HeyangQin committed 2 years ago
  • Add ZeRO++ Japanese blog (#3797)
    tohtana committed 2 years ago
  • Bug Fixes for autotuner and flops profiler (#1880)
    cli99 committed 2 years ago
  • Missing strided copy for gated MLP (#3788)
    cmikeh2 committed 2 years ago
  • Requires grad checking. (#3789)
    jomayeri committed 2 years ago
  • bump to 0.10.0
    jeffra committed 2 years ago
  • Fix Bug in transform.cu (#3534)
    rraminen committed 2 years ago
  • bug fix: triton importing error (#3799)
    stephen-youn committed 2 years ago
  • Merge branch 'master' into revert-3166
    jeffra committed 2 years ago
  • Merge branch 'master' into revert-3166
    jeffra committed 2 years ago
  • Merge branch 'master' into revert-3166
    tjruwase committed 2 years ago
  • Merge branch 'master' into revert-3166
    loadams committed 2 years ago
  • Merge branch 'master' into revert-3166
    tjruwase committed 2 years ago
  • Merge branch 'master' into revert-3166
    tjruwase committed 2 years ago
  • Merge branch 'master' into revert-3166
    tjruwase committed 2 years ago
  • Merge branch 'master' into revert-3166
    tjruwase committed 2 years ago
FAQ Terms Privacy Refunds Impressum
Loading