All pull requests

Pull requests

• Add s390x for llama model

  ready
  #1052·nikheal2 opened Sep 24, 2026Collaborator·

  ·
  1

• chore: bumps vllm to v0.28.0

  #1051·Daniel-Schenker opened Sep 22, 2026Collaborator·

  ·
  5

• [Feature]: Cast model outputs to fp32 inside the compiled graph

  #1049·jan-hofmeier-ibm opened Sep 11, 2026Contributor·

  ·
  1

• Upgrade torch to 2.13, torchvision to 0.28.0

  #1048·Daniel-Schenker opened Sep 3, 2026Collaborator·

  ·
  1

• prefix caching for MM

  #1039·nikheal2 opened Jul 26, 2026Collaborator·

  ·
  22

• Experimental branch, do not merge

  #1000·maxdebayser opened Jun 8, 2026Collaborator·

  ·
  1

• Introducing Sim mode

  #990·yannicks1 opened Jun 3, 2026Collaborator·

  ·
  24

• [WIP] rewrite volumetric constraint check

  #978·yannicks1 opened May 26, 2026Collaborator·

  ·
  6

• [WIP] scheduler exploration sort based on abs tkv difference

  #971·yannicks1 opened May 21, 2026Collaborator·

  ·
  4

• [WIP] scheduler exploration: sort the waiting queue

  #967·yannicks1 opened May 20, 2026Collaborator·

  ·
  4

• [WIP] new scheduler policy based on tkv shift ratio

  #957·yannicks1 opened May 4, 2026Collaborator·

  ·
  3

• [WIP] Scheduler experimentation: baseline scheduler on main

  #907·yannicks1 opened Apr 9, 2026Collaborator·

  ·
  56

• [WIP] Scheduler experimentation: don't shift tkv for new sequence

  #797·yannicks1 opened Mar 6, 2026Collaborator·

  ·
  20