Bingxu ChenandCursor Agent f113ece5cc Revert "improve: combine vit calls for images from different reqs from one batch (#25910)" (#26442) ...

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

2026-05-29 11:13:34 -07:00

..

scheduler_components

[EPD] Optimize the Mooncake backend (#22587)

2026-05-29 10:42:24 +08:00

async_dynamic_batch_tokenizer.py

[tokenizer] improve non streaming request processing + some small fixes. (#20310)

2026-04-10 15:46:12 -07:00

cache_controller.py

Remove managers' unused fields (#25433)

2026-05-16 09:17:11 +08:00

communicator.py

refactor: extract FanOutCommunicator and use declarative spec table (#22967)

2026-04-16 15:37:19 -07:00

configure_logging.py

[Utils] Make request dump robust to unpicklable server_args and large meta_info (#24767)

2026-05-09 21:41:41 -07:00

data_parallel_controller.py

Bundle Scheduler rank/size fields into a frozen ParallelState (#25444)

2026-05-16 09:23:51 +08:00

detokenizer_manager.py

perf: migrate Req token-id storage to array.array('q') in Scheduler (#25098)

2026-05-22 10:51:07 -07:00

disagg_service.py

[PD] Cleanup BootstrapServer init and ready check (#19551)

2026-02-28 16:41:42 +08:00

embed_types.py

[Score API] Hoist query placeholder scan and specialize PositionalEmbeds stacking (#23513)

2026-04-29 13:51:53 -07:00

hisparse_coordinator.py

[core] Maintain req_pool_indices_cpu host mirror (like seq_lens_cpu) (#26425)

2026-05-26 18:59:54 -07:00

io_struct.py

[EPD] Optimize the Mooncake backend (#22587)

2026-05-29 10:42:24 +08:00

mm_utils.py

Revert "improve: combine vit calls for images from different reqs from one batch (#25910)" (#26442)

2026-05-29 11:13:34 -07:00

multi_tokenizer_mixin.py

[BugFix] preserve cached token details in multi-tokenizer output (#26590)

2026-05-28 13:31:55 -07:00

multimodal_processor.py

[Feature] Stronger transformers modeling backend with TP, PP, MoE, VLMs, and torch compile (#19163)

2026-04-02 16:02:33 -07:00

overlap_utils.py

[core] Make overlap-schedule WAR barrier CUDA-only (#26646)

2026-05-29 01:02:31 -07:00

prefill_delayer.py

[PrefillDelayer] support NCCL all-gather for cross-DP info sync (#24768)

2026-05-09 21:20:03 -07:00

schedule_batch.py

Revert "improve: combine vit calls for images from different reqs from one batch (#25910)" (#26442)

2026-05-29 11:13:34 -07:00

schedule_policy.py

Support batch size > 1 when enable CP (#23269)

2026-05-27 14:11:17 -07:00

scheduler_input_blocker.py

Tiny move files to utils folder (#11166)

2025-10-03 22:40:06 +08:00

scheduler_pp_mixin.py

[PP] Skip PP output communication for pure chunked prefill batches (#26148)

2026-05-26 21:59:18 +08:00

scheduler_recv_skipper.py

[Feat]Add scheduler recv skipper weights to environment configuration (#13855)

2025-11-27 18:16:11 +08:00

scheduler.py

[core] Make overlap-schedule WAR barrier CUDA-only (#26646)

2026-05-29 01:02:31 -07:00

template_detection.py

Clean up server startup log noise (#26205)

2026-05-24 14:35:15 -07:00

template_manager.py

Clean up server startup log noise (#26205)

2026-05-24 14:35:15 -07:00

tokenizer_control_mixin.py

Support getting checksums in weight checker (#24537)

2026-05-06 22:59:28 +08:00

tokenizer_manager_score_mixin.py

Replace single-line defensive getattrs with direct access (#25435)

2026-05-16 09:18:43 +08:00

tokenizer_manager.py

[EPD] Optimize the Mooncake backend (#22587)

2026-05-29 10:42:24 +08:00

tp_worker.py

[core] step 1: route non-spec seq_lens via FutureMap with per-mode bootstrap fixes (#25944)

2026-05-21 20:15:51 -07:00

utils.py

[SPEC] fix: use effective max draft tokens for adaptive spec initiali… (#26354)

2026-05-28 13:33:11 -07:00