Commit Graph
518 Commits
Author SHA1 Message Date
fzyzcjy b12b40dea4 Add cuda_graph_forward_passes_total and num_retracted_reqs_total (#15189) 2025-12-17 21:19:42 +08:00
Lianmin Zheng 9d64a7b24f Minor style fixes to the scheduler.py (#15218) 2025-12-16 17:09:44 -08:00
amysaq2023 ccc8f3b266 support non disturbing remote instance weight loader v2 (#14997)
Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
2025-12-16 14:39:56 -08:00
Hanming Lu e61dabf5e4 [Qwen3-next] support mamba radix cache for overlap scheduler (#14792) 2025-12-14 18:54:16 -08:00
9acb21ae27 feat: support EPD disaggregation (#12263)
Co-authored-by: liusy58 <liusy58@linux.alibaba.com>
Co-authored-by: ZhengWG <zwg0606@gmail.com>
Co-authored-by: Nicholas <45984215+liusy58@users.noreply.github.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
2025-12-14 22:30:08 +08:00
fzyzcjy 168a31eb00 Support prefill max requests limitation (#14993) 2025-12-14 10:52:10 +08:00
liupeng374 3134d2b2a7 [scheduler] enhance scheduler in dp_attention mixed case with spec (#14201) 2025-12-14 02:52:26 +08:00
Liangsheng Yin ed52d01b0b Fix spec info's filter when reqs are finished right after prefill (#14742) 2025-12-14 00:32:54 +08:00
Liangsheng Yin 01e3b3f3a3 Fix decode OOM caused by retraction (#14939) 2025-12-13 12:59:17 +08:00
fzyzcjy 313f59ad80 Add soft watchdogs to debug soft hangs (#15023) 2025-12-13 10:41:35 +08:00
fzyzcjy 487cf81a68 Tiny extract SchedulerWatchdog (#15021) 2025-12-13 10:40:32 +08:00
Chenxi Li 9b9d21312a Feature/Fix multi lora scheduler blocking issue and evict LoRA None lastly (#14795) 2025-12-12 17:13:05 -08:00
Yineng Zhangandfzyzcjy 4b7b5af36a Revert several PRs (#14958)
Co-authored-by: fzyzcjy <ch271828n@outlook.com>
2025-12-12 11:25:12 -08:00
Kevin LiandShangming Cai 8fa8d9d7e8 [PD] Add decode PP event loop for PD disaggregation (#14945)
Signed-off-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
2025-12-12 20:15:00 +08:00
c01b2ee094 [PP] Refactor PP to async mode (#11852)
Signed-off-by: Shangming Cai <csmthu@gmail.com>
Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: bluecoffee8 <jasperli2002@gmail.com>
Co-authored-by: zhangxiaolei123456 <zhangxiaolei.666@bytedance.com>
Co-authored-by: ybyang <10629930+whybeyoung@users.noreply.github.com>
2025-12-12 12:54:16 +08:00
Xinyuan Tong 9975acf50f [refactor] Update reasoning parameter to require_reasoning (#14922)
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
2025-12-11 16:55:59 -08:00
amysaq2023 70758d457e support non-disturbing remote-instance-weight-loader (#13125)
Signed-off-by: Anqi Shen <amy.saq@antgroup.com>
2025-12-11 16:45:32 -08:00
ShawnY112358and赵晨阳 f077436831 [fix] Fix issues for in-flight weight updates (#14064)
Co-authored-by: 赵晨阳 <zhaochen20@outlook.com>
2025-12-09 18:12:19 -08:00
b8zhong 55504df2f7 Add FP8 Blockwise GEMM Backend Flag --fp8-gemm-backend (#14379) 2025-12-09 12:05:56 -08:00
Muqi LiandXinyuan Tong 06836ad02a [Reasoning + Structured Output] make reasoning compatible with structured output (#12551)
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
2025-12-08 01:28:39 -08:00
roikoren755 2ce121a1c3 Enable RadixCache for Mamba2 models (#13584) 2025-12-05 18:23:58 +08:00
Lianmin Zhenggithub-actions[bot] <github-actions[bot]@users.noreply.github.com>Hanming LuHanming Lu
64092c8b55 [Auto Sync] Rename is_hybrid to is_hybrid_swa (#14252)
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Hanming Lu <69857889+hanming-lu@users.noreply.github.com>
Co-authored-by: Hanming Lu <hanming@x.ai>
2025-12-01 23:24:24 -08:00
liupeng374andEven Zhou 2e8f54e61e [spec-overlap] bugfix for pd disaggregation and npu (#14088)
Co-authored-by: Even Zhou <even.y.zhou@outlook.com>
2025-12-01 22:58:20 +08:00
fzyzcjy 340c613ab5 Support numactl bind for CPU and memory before process starts (#14156) 2025-11-30 17:00:33 +08:00
fzyzcjy 45cf575852 Fix overlap scheduler not take effect when outputing logprobs (#14096) 2025-11-28 18:15:56 +08:00
ab9a46d462 Support configuring the request limit per receiving poll (#14076)
Co-authored-by: Peng Wang <peng_wang@linux.alibaba.com>
Co-authored-by: Feng Su <225349073+sufeng-buaa@users.noreply.github.com>
2025-11-28 16:14:21 +08:00
Yixin DongandLiangsheng Yin 6350042696 feat: Naive support Spec V2 + Constrained Decoding (#13425)
Signed-off-by: Ubospica <ubospica@gmail.com>
Co-authored-by: Liangsheng Yin <lsyincs@gmail.com>
2025-11-27 20:31:46 +08:00
fzyzcjy 25758647b1 Support sanity checking weight consistency especially for RL (#13854) 2025-11-27 20:25:12 +08:00
Liangsheng Yin eff6a07c8f Fix get_load API (#13991) 2025-11-26 22:07:51 +08:00
vipwangerxiaoandPeng Wang 15729dbc8e Use dynamically maintained num_waiting_tokens in get_load() (#13203)
Signed-off-by: Peng Wang <peng_wang@linux.alibaba.com>
Co-authored-by: Peng Wang <peng_wang@linux.alibaba.com>
2025-11-26 20:34:45 +08:00
Zehuan LiandTiwei Bie 21b0582d4b [feature] Initial block diffusion language model support (#12588)
Co-authored-by: Tiwei Bie <tiwei.btw@antgroup.com>
2025-11-26 17:57:54 +08:00
ShawnY112358and赵晨阳 007c3e234c [feat] support in-flight weight update (#10071)
Co-authored-by: 赵晨阳 <zhaochen20@outlook.com>
2025-11-25 22:03:13 -08:00
ant-yy a146f833f1 [Fix]: Adjust FutureMap's token_id_bufs Size to Prevent ChunkedPrefill's next_token_ids from Overwriting Previous Prefill Requests' next_token_id (#13713)
Signed-off-by: vito.yy <vito.yy@antgroup.com>
2025-11-25 01:08:52 +08:00
Liangsheng Yin b2f7b08c49 Refactor cache init logic (#13800) 2025-11-24 11:41:46 +08:00
Liangsheng Yin 5c2915494c [Scheduler] Tiny organize code style (#13806) 2025-11-23 23:41:34 +08:00
Liangsheng Yin 863124684c [Spec v2] Remove allocate_lens and enable over-allocation (#13478) 2025-11-22 22:49:10 +08:00
Liangsheng Yin 3e804bb0ba Update mem checker during busy (#13704) 2025-11-22 21:15:02 +08:00
b30f63c40f [Bugfix] Fix hidden state size in EAGLE PD disaggregation buffers (#13590)
Co-authored-by: ZeldaHuang <hzm414167@alibaba-inc.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
2025-11-21 15:32:40 +08:00
Even Zhou 2aec8b6e1b [Feature] Spec-Overlap supporting DP-ATTN; PD-Disaggregation; npugraph mode (#12443) 2025-11-15 21:51:07 +08:00
Yuan Luoandluoyuan.luo f0021c0dc8 Add feature flag for mm inputs processing optimization (#13278)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
2025-11-15 14:43:44 +08:00
dtc 8779100768 Fix wrong running_bs in priority scheduling (#13142)
Signed-off-by: Tianchen Ding <dtcccc@linux.alibaba.com>
2025-11-14 03:56:38 +08:00
706502ff6c [VLM] Support PP for Qwen2.5-VL (#13075)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
Co-authored-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
2025-11-12 23:18:44 +08:00
Charlie Ruan 2864c49fbc [RPC] Fix handle_rpc_request with **recv_req.parameters (#7906) 2025-11-12 19:29:12 +08:00
Yikai Zhang 9c546bfd6d Fix the Wrong Return Type of Scheduler.recv_requests (#7886) 2025-11-12 18:54:56 +08:00
wxsm f825137f8a fix: Remove dupulicated kv_events initialization in scheduler (#13132) 2025-11-11 23:54:47 -08:00
vipwangerxiaoandPeng Wang 8f01a12d43 Improve overlap scheduling for better TTFT (#11856)
Co-authored-by: Peng Wang <peng_wang@linux.alibaba.com>
2025-11-12 11:24:04 +08:00
yinghui 38a704bccb refine stdout logging codes (#13015) 2025-11-12 00:16:14 +08:00
Liangsheng Yin f09eee036d Tiny simplify evcition metrics collector (#12983) 2025-11-11 23:23:48 +08:00
Feng Su fc8cda14cf Sglang Tracing: optimize trace_event_batch() (#13036)
Signed-off-by: Feng Su <sufeng@linux.alibaba.com>
2025-11-11 15:02:20 +08:00
Liangsheng Yin 665416f6dd Unify memory management across (overlap, non-overlap) x (page>=1) x (spec, non-spec, spec v2) x (retract, finished) (#12224) 2025-11-11 02:56:22 +08:00