This website requires JavaScript.
Explore
Help
Register
Sign In
minke.yu
/
sglang
Watch
1
Star
0
Fork
0
Code
Issues
Pull Requests
Actions
12
Packages
Projects
Releases
Wiki
Activity
All Workflows
Build and Publish Multi-Arch Docker Images
Cleanup Old Nightly Docker Tags
PR Test Stage for NPU
Single Node Template for E2E performance and accuracy tests
Check Changes
PR Test - Build Rust Extensions
PR Test - Build sgl-kernel Wheels
PR Test SGLang Simulator (CPU)
_pr-test-simulator-gpu.yml
PR Test Stage (CPU)
_pr-test-stage.yml
AMD AITER Scout
AMD CI Job Monitor
Auto tune
Bot Bump Docs Version
Bot Bump Flashinfer Version
Bot Bump Kernel Version to SGLang
Bot Bump Kernel Version
Bot Bump SGLang Version
Bot Cherry Pick to Release Branch
Cancel PR Workflows on Close
Cancel Unfinished PR Runs
CI Auto Bisect
CI Coverage Overview
CI Failure Monitor
CI Lark Notify
CI Model Inventory
Close Inactive Issues
Close Stale PRs
Coverage Collection (NPU)
Diffusion CI Ground Truth Generation (NPU)
Diffusion CI Ground Truth Generation
Full Test (NPU)
Auto Label PRs
Lint
List Active Runs
MLU CI Reliability Report
Nightly Test (GB200 72GPU)
Nightly Test (AMD MI355X Disagg)
Nightly Link Check
Nightly Release SGLang Model Gateway to PyPI
Nightly sgl-router Docker Image
Nightly Test Miles (AMD ROCm 10 MI30X)
Nightly Test Miles (AMD ROCm 10)
Nightly Test (AMD)
Nightly Wheel Test (Intel)
Nightly Main Test (Intel)
Nightly Test (MLU)
Nightly Test (MUSA)
Nightly Test (NPU) Multi Node Template
Nightly Test (NPU)
Nightly Test (Nvidia)
Open A PR to Copy Code From OSS
Open A PR to Copy Diff To OSS
Patch Docker Image
PR Benchmark (SMG Components)
PR Check Comment
pr-gate.yml
PR States
PR Test Extra (AMD)
PR Test Mori (AMD)
PR Test (AMD)
PR Test (Arm64)
PR Test Extra
pr-test-jit-kernel.yml
PR Test (MLU)
PR Test (MLX)
pr-test-multimodal-gen.yml
PR Test (MUSA)
PR Test (NPU)
PR Test (PPU)
PR Test (Rust extensions)
PR Test (Rust workspace)
PR Test (SMG)
pr-test-sgl-kernel.yml
PR Test (sglang-processor)
PR Test (sgl-router)
PR Test (Xeon)
PR Test (XPU)
PR Test Base
PR Title
Release Branch Cut
Release Docker Images Nightly gfx1151 (AMD)
Release Docker Images Nightly Miles ROCm10 (AMD)
Release Docker Images Nightly Miles ROCm7.2.4 (AMD)
Release Docker Images Nightly ROCm 10 (AMD)
Release Docker Images Nightly ROCm7.2 (AMD)
Release Docker Images (AMD)
Release Docker Images Nightly (CUDA 13.4 Rubin)
Build and Push Development Docker Images
Release Docker EFA Images
Release SGLang Model Gateway Docker Image
Release Docker Images Nightly (Intel XPU)
Release Docker Images (Intel XPU)
Release Docker Images Nightly (NPU)
Release Docker Images (NPU)
Release Docker Runtime Images
Release Docker Xeon Images
Release Docker Images
Release SGLang Model Gateway to PyPI
Release PyPI Nightly Wheels
Release PyPI PR Wheels
Release PyPI
Release Tag
Release sgl-deep-ep
Release sgl-deep-gemm
Release sgl-flash-mla
Release SGLang XPU Kernel
Release SGLang Kernels
Rerun Test
Retag Docker Image
Run CPU CI
Runner Utilization Report
Seed Rust Ext Cache
Slash Command Handler
Stress Test
Sync LMSYS SGLang blogs
Trivy Scan Dev Docker Images
Weekly Test (CPU)
Weekly Test (Nvidia)
Weekly Update Est Time
XPU CI Job Monitor
Other workflows
2
build-image.yaml
release-docker-amd-miles-rocm720-nightly.yml
31 workflow runs
Actor
All actors
injet
minke.yu
Status
All status
Success
Failure
Canceled
Skipped
Waiting
Running
Blocked
Canceling
Branch
All branches
b300/dynamo
b300/integration
b300-pick-261102
b300-pick-261102-v3
b300-pick-46884-async-20261107
bot/update-est-time-20260928
bot/update-est-time-20261005
bot/update-est-time-20261012
bot/update-est-time-20261019
bot/update-est-time-20261026
bot/update-est-time-20261102
bot/update-est-time-20261109
bot/update-est-time-20261116
dsv41-pd
dsv41-pd-visioncp
g9i-est-td
gist-nvlink
gw-ns-est-td-20261115
gw-ns-g9i-20261114
gw-schema-compat
h20-est
main
Create status badge
G9 P0 fixes: v2 pool_transfers for storage hit query, disagg drain hook, query gate
#400
:
Commit
e7dc175e94
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-11 18:08:50 +08:00
21m51s
View workflow file
hicache L3 prefetch: io_aux 传输并行化(按 op 分 lane)
#381
:
Commit
baba18bc7d
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 19:53:52 +08:00
48m34s
View workflow file
hicache: STORAGE_BATCH_SIZE 可env化(SGLANG_HICACHE_STORAGE_BATCH_SIZE,默认128不变)
#378
:
Commit
7d5759b2d9
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 18:14:15 +08:00
18m40s
View workflow file
mooncake_store: local_hostname 加 rank 后缀,让 master 的 preferred-segment 分配落到本进程段
#377
:
Commit
84fc632088
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 16:42:09 +08:00
1s
View workflow file
hicache: SGLANG_HICACHE_TIMING 打点 L3 预取全链路分阶段耗时
#375
:
Commit
5f2ffa70b8
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 15:29:21 +08:00
2s
View workflow file
测试适配
#38348
新语义:SWA 双层缺失的 backuped 节点现在是合法匹配边界
#369
:
Commit
a42ab6fce0
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 07:46:24 +08:00
1s
View workflow file
手工移植上游 #38348:hicache SWA prefetch fail-soft 降级
#368
:
Commit
fa8ad18b5c
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 07:23:50 +08:00
1s
View workflow file
cherry-pick 上游 #40429:FlashMLA sched meta 超 48KiB smem 时跳过预计算(encoder/decoder replay 的大 batch replay pass 会触发 smem 溢出崩溃)
#366
:
Commit
7f1edbba80
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 01:20:26 +08:00
1s
View workflow file
cherry-pick 上游 #42716(运行时部分):encoder SWA bounded replay 下放行 host 层前缀命中——request_window 形态的 SWA 不进树,校验门不应拦截 host-only 节点;rust adapter/test 两处 hunk 与本树结构不符未合(默认 python 树芯用不到)
#365
:
Commit
99c317b131
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 01:02:42 +08:00
2s
View workflow file
mooncake standalone(dummy) 模式打通 dsv41 unified cache:放宽 standalone 检查对 LogicalHostPool 锚点的误拦(实体 buffer 由旁池运行时 register_buffer);新增 l3dummy compose(单 store 服务 64GB shm 段 + fake-cargo 跳过 rustup 下载)。实测 24x160k 热/冷 …
#364
:
Commit
b4dbcebbec
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-10 01:17:16 +08:00
1h10m35s
View workflow file
l3test: prefetch 策略 wait_complete→best_effort(实测等待阻塞净亏 7-17%,best_effort 持平 -4.4%)
#361
:
Commit
30ba5818e7
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 20:31:01 +08:00
1s
View workflow file
revert:
#42035
continuous prefill input polling (CP rank collective mismatch deadlock)
#351
:
Commit
7578cdd218
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 17:18:29 +08:00
39m2s
View workflow file
build: 镜像内编译 sgl-model-gateway(Rust gateway,G9 前置)
#350
:
Commit
21f7ce37dd
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 16:39:39 +08:00
31m52s
View workflow file
test: adopt upstream
#42035
auxiliary-output asserts now that checkpoint_disagg_prefill shim exists
#347
:
Commit
1f6ca2cf68
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 12:02:22 +08:00
26m16s
View workflow file
[mem_cache] Skip the release-time insert on optimistic prefill requeue; make
refresh_fill_ids
public (
#42194
)
#346
:
Commit
9ef65b51c2
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 11:44:37 +08:00
25m53s
View workflow file
test: adapt
#42035
polling test to pre-#41520 name maybe_cache_unfinished_req
#345
:
Commit
6ba34507f1
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 11:37:08 +08:00
26m20s
View workflow file
[PD] Keep ingesting requests while a prefill forward result is pending (
#42035
)
#344
:
Commit
93d610c1fd
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-09 11:26:55 +08:00
26m6s
View workflow file
fix: import is_hip_runtime in dsv4 low_ratio_compress (cherry-pick
#41660
lost it)
#327
:
Commit
1068926cd6
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-08 17:04:05 +08:00
32m19s
View workflow file
build: v3 full path reuses base torch/nvidia via local delta check
#325
:
Commit
8d0ee5e51e
pushed by
minke.yu
dsv41-pd-visioncp
2026-10-08 16:23:59 +08:00
28m51s
View workflow file
scheduler: CP-symmetric idle check for health-check admission
#65
:
Commit
b91137ab98
pushed by
minke.yu
dsv41-pd-visioncp
2026-09-24 12:30:02 +08:00
27m52s
View workflow file
fix: vision check uses get_parallel().pp_group (get_pp_group undefined in this tree)
#49
:
Commit
db7d2cb7db
pushed by
minke.yu
dsv41-pd-visioncp
2026-09-23 18:34:19 +08:00
27m58s
View workflow file
更新 .gitea/workflows/build-image.yaml
#47
:
Commit
f23179ce99
pushed by
injet
dsv41-pd-visioncp
2026-09-23 18:21:21 +08:00
26m23s
View workflow file
更新 .gitea/workflows/build-image.yaml
#46
:
Commit
8266769b2d
pushed by
injet
dsv41-pd-visioncp
2026-09-23 17:50:46 +08:00
1m12s
View workflow file
ci: also trigger image build on dsv41-pd-visioncp pushes
#45
:
Commit
7925735a3e
pushed by
minke.yu
dsv41-pd-visioncp
2026-09-23 17:16:01 +08:00
29m50s
View workflow file
update .gitignore
#44
:
Commit
11e5ba6e8f
pushed by
injet
dsv41-pd
2026-09-23 17:11:13 +08:00
28m22s
View workflow file
更新 .gitignore
#42
:
Commit
fb341f8de5
pushed by
injet
dsv41-pd
2026-09-23 16:30:31 +08:00
32m5s
View workflow file
[Fix] Handle chunked paged MQA metadata in DSV4.1 eager forwards (
#40637
)
#40
:
Commit
ddf5207630
pushed by
minke.yu
dsv41-pd
2026-09-23 15:30:23 +08:00
32m15s
View workflow file
ci: tag builds with -ci-<UTC datetime> suffix (naming rule: <branch>-<sha9>-<ci|local>-<UTC>)
#32
:
Commit
104218d9ed
pushed by
minke.yu
dsv41-pd
2026-09-22 21:08:47 +08:00
30m44s
View workflow file
ci: drop gha cache backend (runner cache server 404s on this instance)
#19
:
Commit
92632a60ba
pushed by
minke.yu
dsv41-pd
2026-09-22 19:40:17 +08:00
50m51s
View workflow file
ci: hardcode Gitea registry (vars context falls back to docker.io on this instance)
#16
:
Commit
4b3b367b63
pushed by
minke.yu
dsv41-pd
2026-09-22 18:48:03 +08:00
24m13s
View workflow file
First
Previous
1
2
Next
Last
Status Badge
Branch
main
main
b300/dynamo
b300/integration
b300-pick-261102
b300-pick-261102-v3
b300-pick-46884-async-20261107
bot/update-est-time-20260928
bot/update-est-time-20261005
bot/update-est-time-20261012
bot/update-est-time-20261019
bot/update-est-time-20261026
bot/update-est-time-20261102
bot/update-est-time-20261109
bot/update-est-time-20261116
dsv41-pd
dsv41-pd-visioncp
g9i-est-td
gist-nvlink
gw-ns-est-td-20261115
gw-ns-g9i-20261114
gw-schema-compat
h20-est
Badge URL
Markdown
HTML