Default Branch

6376fda7a9 · [Bugfix][Rust Frontend] Use server-reported prompt_tokens in vllm-bench (#59694) · Updated 2026-10-28 09:22:50 +08:00

Branches

e65e4f3a21 · fixes · Updated 2026-10-28 09:07:14 +08:00

43
15

80b23fc460 · [MiniMax-M3.1] Fix CI: lazy CuTe DSL imports, CUDA 13.0 skips, test configs · Updated 2026-10-28 05:10:22 +08:00

23
7

d0f0b04761 · Log skipped and failed DeepEP GIN barrier patch application · Updated 2026-10-28 03:47:15 +08:00

23
5

a0ea941097 · [Bugfix][MiniMax-M3] Scope prompt reasoning-end checks to the last message (#63492) · Updated 2026-10-28 00:30:14 +08:00

97
3

56ea7961fc · remove dead code multiple places · Updated 2026-10-27 23:31:15 +08:00

44
1

a0fe115ac9 · [CI/Build] Publish nightly EFA image to DockerHub · Updated 2026-10-27 19:04:20 +08:00

60
1

7863c62d1f · [CI] Re-enable disabled jobs whose installs are already locked · Updated 2026-10-27 17:28:27 +08:00

80
1

b9ed63353b · Bump astral-sh/setup-uv from 7.6.0 to 10.3.2 · Updated 2026-10-27 12:13:41 +08:00

99
1

da99f5f3f0 · Bump actions/setup-python from 6.1.0 to 7.1.0 · Updated 2026-10-27 12:13:20 +08:00

99
1

682544c6ae · Bump actions/checkout from 7.0.0 to 7.0.1 · Updated 2026-10-27 12:12:58 +08:00

99
1

31f3a6df1f · Merge main; take main's disaggregated and fault-tolerance steps · Updated 2026-10-27 05:49:08 +08:00

120
2

4135590ffe · [Bugfix][MRV2] Skip adaptive verification cost profiling in the CUDA graph memory-profiling pass (#63431) · Updated 2026-10-27 01:59:24 +08:00

138
0
Included

659a55dee1 · Merge branch 'main' into akaratza_enable_ue8m0 · Updated 2026-10-24 05:32:57 +08:00

278
3

16dd7ed4e6 · [CI] Leave GPU headroom in the offline ColBERT plugin tests · Updated 2026-10-24 04:33:27 +08:00

285
1

cfd92bd0f0 · [Core] Remove Model Runner V1 · Updated 2026-10-24 00:53:43 +08:00

300
1

82135ca498 · [Bugfix][CPU] Fix GDN spec conv-state handling for short query lengths · Updated 2026-10-23 21:41:31 +08:00

312
1

e4a181571e · Refactor PCP O-Proj TP into an intermediate PCPOProjTP layer with descriptor-driven weight switching · Updated 2026-10-23 10:43:28 +08:00

349
2

eff8a56896 · [Doc][Recipe] MiniMax-M3.1-Flash: add Hopper FP8 variant, Docker-only install (#63244) · Updated 2026-10-23 10:06:56 +08:00

350
0
Included

81a670767c · [Beam] Enable CuTeDSL NVFP4 MegaMoE · Updated 2026-10-22 06:36:25 +08:00

678
6

817fa3518c · [Frontend] Add derender output mode to generate (#60082) · Updated 2026-10-21 23:20:46 +08:00

461
0
Included