Code

Update llama examples#90

Code

Merged

IlyasMoutawwakil merged 29 commits intomainhuggingface/optimum-benchmark:mainfrom update-llama-exampleshuggingface/optimum-benchmark:update-llama-examplesCopy head branch name to clipboardNov 30, 2023

+744-4,298

Commits

Commits on Nov 24, 2023

• added new configs

  IlyasMoutawwakil

  committed
  5668117

• update

  IlyasMoutawwakil

  committed
  5c67276

• fix

  IlyasMoutawwakil

  committed
  3f37606

Commits on Nov 26, 2023

• added decode latency and throughput

  IlyasMoutawwakil

  committed
  038abd3

• fix dp config

  IlyasMoutawwakil

  committed
  387d012

• fix decode metrics

  IlyasMoutawwakil

  committed
  f412d81

• added a100 results

  IlyasMoutawwakil

  committed
  4dc5b95

Commits on Nov 27, 2023

• fix distributed throughput measurments

  IlyasMoutawwakil

  committed
  e194f6b

• fix report

  IlyasMoutawwakil

  committed
  ce60f97

• added fp16+tp=1 for deepspeed baseline

  IlyasMoutawwakil

  committed
  23801a1

• make isolation process failure more verbose

  IlyasMoutawwakil

  committed
  bf5626d

• update A100 results

  IlyasMoutawwakil

  committed
  adbc1e9

• added ddp config

  IlyasMoutawwakil

  committed
  8aec23b

• update custom reporting script

  IlyasMoutawwakil

  committed
  225d165

• remove tp=1

  IlyasMoutawwakil

  committed
  301c608

• style

  IlyasMoutawwakil

  committed
  3ab2533

• added bt and fa2 configs

  IlyasMoutawwakil

  committed
  cccced8

• fix isolation process

  IlyasMoutawwakil

  committed
  6b6513c

• added gptq config

  IlyasMoutawwakil

  committed
  14a32a6

Commits on Nov 28, 2023

• update configs

  IlyasMoutawwakil

  committed
  e61ca6b

• more isolation process changes

  IlyasMoutawwakil

  committed
  00543de

• added an inline launcher (as default)

  IlyasMoutawwakil

  committed
  c27c708

• update distributed configs

  IlyasMoutawwakil

  committed
  f00d042

• fix

  IlyasMoutawwakil

  committed
  08bc26b

• added fa2+tp=2 config and added llama 65B gptq

  IlyasMoutawwakil

  committed
  2584fd2

• disable no_weights because of of 65B model error

  IlyasMoutawwakil

  committed
  6ddfdf7

• added all A100 GPTQ results

  IlyasMoutawwakil

  committed
  873f799

Commits on Nov 29, 2023

• added A100 reports for llama 7b, 13b and 65b

  IlyasMoutawwakil

  committed
  c8bbe4e

• remove experiments (only leave artifacts)

  IlyasMoutawwakil

  committed
  370d7ae