All pull requests

Pull requests

• [BugFix][Speculative Decoding] MTP logprobs path drops enable_pd_reorder and raises TypeError

  #8120·Anai-Guo opened Aug 30, 2026·

  ·
  1

• [BugFix][GCU] post_process call sites still use the renamed sampler_output kwarg

  #8119·Anai-Guo opened Aug 30, 2026·

  ·
  1

• 【Hackathon 10th Spring No.50】为 FastDeploy 新增 MiniCPM4.1-8B 模型

  #8118·codecason opened Aug 28, 2026·

  ·
  1

• [XPU] Fix PaddleFormers fallback model issues

  #8116·mayang002 opened Aug 25, 2026Contributor·

  ·

• [Benchmark] support session control in multi turn test

  #8106·liyonghua0910 opened Aug 6, 2026Collaborator·

  ·

• optimize get_rdma_nics.sh to support nsys profile

  #8099·wangyuwen1999 opened Jul 13, 2026Contributor·

  ·
  6

• [BugFix] Fix image_features is None crash in PaddleOCR-VL

  contributor
  #8096·nandanadileep opened Jul 7, 2026·

  ·
  1

  7

• [RL] enforce cuda graph clearing and rebuilding in rsync weight update by rdma

  #8085·liyonghua0910 opened Jun 29, 2026Collaborator·

  ·
  4

• [Cherry-Pick][Others] support arm env

  #8084·BingooYang opened Jun 29, 2026Contributor·

  ·
  14

• [Feature]report PD info to IM

  #8082·ChowMingSing opened Jun 26, 2026Contributor·

  ·
  6

• [Cherry-pick][PD Disaggregation] Limit prefill fetch num with FD_MAX_INFLIGHT_PREFILL(#7981)

  #8081·Sunny-bot1 opened Jun 26, 2026Collaborator·

  ·
  5

• [RL] Reuse GDR checkpoint transfer handle

  #8078·jackyYang6 opened Jun 25, 2026Contributor·

  ·
  8

• Fix MLA cache shape init without layerid

  #8072·Socratesa opened Jun 23, 2026·

  ·
  5

• [XPU] add fetch req log

  #8070·cmcamdy opened Jun 23, 2026Collaborator·

  ·
  8

• feat(metax): SigLIP vision encoder optimizations (+24% throughput)

  #8068·valorix25 opened Jun 22, 2026·

  ·
  5

• [Cherry-Pick][BugFix] Seperate prometheus multiproc dir for single-server multi-dp services (#8059)

  #8062·liyonghua0910 opened Jun 17, 2026Collaborator·

  ·
  14

• [BugFix] Seperate prometheus multiproc dir for single-server multi-dp services

  #8059·liyonghua0910 opened Jun 16, 2026Collaborator·

  ·
  21

• swa

  #8054·zhoutianzi666 opened Jun 16, 2026Collaborator·

  ·
  5

• Support swa mha

  #8053·zhoutianzi666 opened Jun 16, 2026Collaborator·

  ·
  6

• [Refactor] Remove redundant code in MLA cache management

  #8050·HayzelHan opened Jun 15, 2026·

  ·
  6

• [Models] Support MLA_SWA functionality

  #8049·chang-wenbin opened Jun 15, 2026Collaborator·

  ·
  10

• clean code

  #8044·zhoutianzi666 opened Jun 15, 2026Collaborator·

  ·
  6

• [XPU] fix_same_req_id

  #8040·cmcamdy opened Jun 11, 2026Collaborator·

  ·
  5

• Diff kv head

  #8037·chang-wenbin opened Jun 10, 2026Collaborator·

  ·
  15

• fp8 triton moe config.

  #8036·xuanyuanminzheng opened Jun 10, 2026Collaborator·

  ·
  3