Repository navigation

Fix moe acc - #7988

Open

BingooYang wants to merge 143 commits into

PaddlePaddle:developfrom

BingooYang:fix_moe_acc

Commits on Apr 3, 2026
Commits on Apr 8, 2026
- [Cherry-Pick][BugFix] fix MTP bugs in TP and overlap(#7172) (#7192)
  huicongyao
  authored
  36909bf
- remove arctic_inference deps (#7236)
  Deleter-D
  authored
  403ce13
- Split enable_mm (#7183) (#7233)
  
  authored
  6b78981
- [Feature]distinguish whl version (#7204) (#7224)
  
  EmmonsCurse
  and
  ChowMingSing
  authored
  84d6271
- support moe for sm103 (#7240)
  BingooYang
  authored
  0181884
Commits on Apr 9, 2026
Commits on Apr 10, 2026
Commits on Apr 11, 2026
Commits on Apr 12, 2026
- [Cherry-Pick][CI] Sync dev optimizations to 2.6(#7335) (#7343)
  EmmonsCurse
  authored
  9e8ea7d
Commits on Apr 13, 2026
- [Cherry-Pick][TI-consistent] support quant use pow2scale(#7308) (#7310)
  liuruyan
  authored
  9cb82d7
- fix overlap mtp empty run (#7314)
  Sunny-bot1
  authored
  b2997f3
- [Feature] Support set PREEMPTED_TOKEN_ID in GET_SAVE_OUTPUT_V1 (#7159) (#7351)
  rainyfly
  authored
  d9a008f
- remove fa4 requirements (#7354)
  zoooo0820
  authored
  9823d63
- update attn_mask_q 2 (#7373)
  ckl117
  authored
  144dc17
Commits on Apr 14, 2026
Commits on Apr 15, 2026
- check init_flash_attn_version log (#7401)
  ckl117
  authored
  2ee1cc3
- modify flashmask version (#7414)
  BingooYang
  authored
  61bfe6e
- [Cherry-Pick][RL] Add clear_graph_opt_backend for glm4_mtp (#7378) (#7379)
  Deleter-D
  authored
  26674bb
Commits on Apr 16, 2026
- PD deployment support without router (#7412) (#7424)
  juncaipeng
  authored
  b8e8a62
- [BugFix] fix tool call parser (#7369) (#7419)
  
  EmmonsCurse
  and
  luukunn
  authored
  72ce56b
Commits on Apr 17, 2026
Commits on Apr 20, 2026
- [Bugfix][RL] fix control request timeout in async update weights pipeline (#7470)
  jackyYang6
  authored
  fc801f8
- [CI] Temporarily pin paddlepaddle-gpu to 3.5.0.dev20260417 (#7486) (#7519)
  EmmonsCurse
  authored
  f4f7760
Commits on Apr 21, 2026
Commits on Apr 22, 2026
Commits on Apr 23, 2026
- support deepgemm without bias input (#7559) (#7565)
  
  EmmonsCurse
  and
  zoooo0820
  authored
  258b22a
- [KSM] support keep sampling mask (#7460)
  zeroRains
  authored
  b3aa469
- [Cherry-Pick][BugFix] Fix save_output_specualate parameter bugs in suffix decoding (#7566) (#7569)
  Deleter-D
  authored
  eb92613
- Cache queue support ipc (#7589)
  juncaipeng
  authored
  10f5a20
Commits on Apr 24, 2026
Commits on Apr 27, 2026
Commits on Apr 28, 2026
Commits on Apr 29, 2026
Commits on Apr 30, 2026
Commits on May 1, 2026
- [BugFix] Fix get_tasks returns empty list and incorrect nnode computation (additional fixes) (#7685)
  liyonghua0910
  authored
  66dea60
Commits on May 8, 2026
- fix rl overlap (#7745)
  Sunny-bot1
  authored
  d0a0b3e
Commits on May 9, 2026
Commits on May 11, 2026
Commits on May 12, 2026
- [BugFix] Fix KSM bug in MTP and Overlap (#7788)
  zeroRains
  authored
  fae4a8b
- [Cherry-Pick][Optimization] enable trtllm_all_reduce fusion kernel in glm model (#6660) (#7228)
  BingooYang
  authored
  ae5dac1
Commits on May 13, 2026
Commits on May 14, 2026
Commits on May 15, 2026
- [Cherry-Pick][BugFix] Fix pause drain hang caused by stale abort markers(#7825) (#7826)
  jackyYang6
  authored
  04e4ae8
- [Cherry-Pick][CI] Optimize clean_ports logic by removing redundant code(#7809) (#7830)
  EmmonsCurse
  authored
  d71bdda
Commits on May 18, 2026
- [Cherry-Pick][Op][Optimization]Kernel fusion: cast+sigmoid+bias+noauxtc(#7777) (#7818)
  ShaneGZhu
  authored
  514ed5c
- [Cherry-Pick][RL] Support cpu tensor broadcast(#7833) (#7840)
  Sunny-bot1
  authored
  9894b32
Commits on May 19, 2026
- [Cherry-Pick][CI] Set --workers=1 to avoid intermittent timeout failures (#7846) (#7848)
  EmmonsCurse
  authored
  ab3c5f4
- fix refact abort (#7838)
  qwes5s5
  authored
  41d44d6
Commits on May 20, 2026
- [Cherry-Pick] update fleet_ops(#7859) (#7858)
  liuruyan
  authored
  8c4f5a6
- [Cherry-Pick][Optimization] Reduce logprob processing overhead by using actual topk instead of fixed K+1 (#7860) (#7861)
  Sunny-bot1
  authored
  31b12ee
Commits on May 21, 2026
- [RL] Reset buffer size of `slot_mapping` (#7868)
  gongshaotian
  authored
  b5c8290
- fix ce bug (#7874)
  liuruyan
  authored
  b562b8d
Commits on May 22, 2026
- [Cherry-Pick][Feature][Log]console metrics log for pd disaggregation #7843 (#7845)
  CSWYF3634076
  authored
  485f6c2
- [Cherry-Pick][Benchmark] Add inner benchmark metrics component (#7881) (#7831)
  Deleter-D
  authored
  e7815be
Commits on May 25, 2026
Commits on May 26, 2026
Commits on May 27, 2026
- [Cherry-Pick][Bugfix] Fix clear bug in RL causing CUDA error 700 during CUDAGraph recapture(#7934) (#7933)
  freeliuzc
  authored
  261041b
- [PD] PD send cache via storage & Refine swap_cache_layout op (#7839)
  juncaipeng
  authored
  8a1e71d
Commits on May 28, 2026
Commits on May 29, 2026
- [Cherry-Pick][BugFix] fix mtp reset bugs in rl (#7957) (#7958)
  Deleter-D
  authored
  ac24fcc
Commits on Jun 1, 2026
- [RL] Fix the incorrect routing of EOS tokens, which leads to changes in accuracy (#7960)
  gongshaotian
  authored
  7198b58
Commits on Jun 2, 2026
- [RL] Fix Ernie mm bug (#7966)
  gongshaotian
  authored
  eeed8a3
Commits on Jun 3, 2026
- [Cherry-Pick][RL][Feature] Add GDR streaming weight update path (#7951) (#7971)
  jackyYang6
  authored
  780c000
- fix moe accurate issue
  BingooYang
  committed
  99c7df1
- fix bug
  BingooYang
  committed
  f232ed9
- add test
  BingooYang
  committed
  b5ec4fa

Commits

Commits on Apr 3, 2026

Commits on Apr 8, 2026

Commits on Apr 9, 2026

Commits on Apr 10, 2026

Commits on Apr 11, 2026

Commits on Apr 12, 2026

Commits on Apr 13, 2026

Commits on Apr 14, 2026

Commits on Apr 15, 2026

Commits on Apr 16, 2026

Commits on Apr 17, 2026

Commits on Apr 20, 2026

Commits on Apr 21, 2026

Commits on Apr 22, 2026

Commits on Apr 23, 2026

Commits on Apr 24, 2026

Commits on Apr 27, 2026

Commits on Apr 28, 2026

Commits on Apr 29, 2026

Commits on Apr 30, 2026

Commits on May 1, 2026

Commits on May 8, 2026

Commits on May 9, 2026

Commits on May 11, 2026

Commits on May 12, 2026

Commits on May 13, 2026

Commits on May 14, 2026

Commits on May 15, 2026

Commits on May 18, 2026

Commits on May 19, 2026