Code

multi-gpu fused_moe tuning support#143

Code

Open

divakar-amd wants to merge 20 commits intomainROCm/vllm:mainfrom distributed_fmoe_tuningROCm/vllm:distributed_fmoe_tuningCopy head branch name to clipboard

+263-238

Commits

Commits on Sep 17, 2024

• add multi-gpu tuning support with tqdm progress bar

  Show description for fd16af5
  divakar-amd

  committed
  fd16af5

• ruff & yapf

  divakar-amd

  committed
  37fc500

• kernel api update & Torchrun usage warning

  divakar-amd

  committed
  6f45d02

• [nit] file mode fix

  divakar-amd

  committed
  ffaffcd

• ruff ruff ruff (isort)

  divakar-amd

  committed
  0e0608f

Commits on Sep 18, 2024

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  86e53fd

Commits on Sep 21, 2024

• fix hardcoded top_k

  divakar-amd

  authored
  fd6a4ee

Commits on Sep 25, 2024

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  ab92950

Commits on Sep 26, 2024

• add exception handling to see silen torchrun failures

  divakar-amd

  committed
  2cb34c4

• yapf

  divakar-amd

  committed
  60860d9

Commits on Oct 14, 2024

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  1e8b1f0

Commits on Oct 16, 2024

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  101564f

• use itertool.product for readability

  divakar-amd

  committed
  618663d

• re-use fused_topk function

  divakar-amd

  committed
  d47b89c

• yapf

  divakar-amd

  committed
  17f9533

• keep the config keys sorted in multi-gpu

  divakar-amd

  committed
  7202911

Commits on Oct 23, 2024

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  dda55c8

• add fp8 tuning support

  divakar-amd

  committed
  c1134b6

• ruff + yapf

  divakar-amd

  committed
  0356fdd

• Merge branch 'main' into distributed_fmoe_tuning

  divakar-amd

  authored
  72dfb5a