DeepSpeed
[inference] ds-attention refactor w.r.t. ops
#2623
Merged
Overview
Commits
17
Changes
View On GitHub
Commits
  • refactor of ds-attn to use new op bindings
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • fix imports
    jeffra committed 3 years ago
  • fix for softmax
    jeffra committed 3 years ago
  • fix typo
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • address comments and consolidate imports
    jeffra committed 3 years ago
  • Merge branch 'jeffra/op-shim' of github.com:microsoft/DeepSpeed into jeffra/op-shim
    jeffra committed 3 years ago
  • fix import issue
    jeffra committed 3 years ago
  • move bloom specific attn into its own class
    jeffra committed 3 years ago
  • remove dead code
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • Merge branch 'master' into jeffra/op-shim
    jeffra committed 3 years ago
  • move softmax op to BloomSelfAttention
    jeffra committed 3 years ago
  • only load op in base to avoid excessive logging
    jeffra committed 3 years ago
FAQ Terms Privacy Refunds Impressum
Loading