Code

Enable cuda graph by default#612

Code

Merged

merrymercy merged 15 commits intomainsgl-project/sglang:mainfrom capturesgl-project/sglang:captureCopy head branch name to clipboardJul 13, 2024

+317-70

Commits

Commits on Jul 13, 2024

• compute matches

  merrymercy

  committed
  e354731

• cuda graph support

  merrymercy

  committed
  74c4545

• fix

  merrymercy

  committed
  517d83a

• pad/unpad

  merrymercy

  committed
  a41861e

• fix padding

  merrymercy

  committed
  283a00a

• update flashinfer version

  merrymercy

  committed
  a20bd0e

• ignore nan in sampling

  merrymercy

  committed
  7c71b56

• fix padding

  merrymercy

  committed
  a3a49f1

• fix padding

  merrymercy

  committed
  f2475b3

• add refernece counter

  merrymercy

  committed
  4e1c870

• update

  merrymercy

  committed
  abdffef

• fix nccl_cumem_enable

  merrymercy

  committed
  6f94576

• update global configs

  merrymercy

  committed
  6da0c3a

• fix bugs

  merrymercy

  committed
  58dfe8c

• fix more bugs

  merrymercy

  committed
  9718caa