Code

[gemini] fixes for benchmarking#5847

Code

Merged

botbw merged 8 commits intomainhpcaitech/ColossalAI:mainfrom gemini_benchhpcaitech/ColossalAI:gemini_benchCopy head branch name to clipboardJun 26, 2024

+27-15

Commits

Commits on Jun 21, 2024

• [gemini] fix missing return

  botbw

  committed
  6370a1b

  Copy full SHA for 6370a1b

• [gemini] fix missing arg pass

  botbw

  committed
  b1b92ba

  Copy full SHA for b1b92ba

• [gemini] use gather tensor instead of list

  botbw

  committed
  eecf9a0

  Copy full SHA for eecf9a0

Commits on Jun 25, 2024

• [test] enable flash attention for benchmark by default

  botbw

  committed
  1e6609a

  Copy full SHA for 1e6609a

• [test] enable flash attention for benchmark by default

  botbw

  committed
  a6b602e

  Copy full SHA for a6b602e

• Merge branch 'gemini_bench' of https://github.com/hpcaitech/ColossalAI into gemini_bench

  Hz188

  committed
  c4d2020

  Copy full SHA for c4d2020

Commits on Jun 26, 2024

• Merge branch 'main' into gemini_bench

  botbw

  authored
  eb62905

  Copy full SHA for eb62905

• [example] use bfloat16 instead with llama benchmark

  botbw

  committed
  498587e

  Copy full SHA for 498587e