[sglang-miles] Fix LoRA usage-counter lifecycle across cleanup paths - #33300

Open

Shi-Dong wants to merge 44 commits into

sgl-project:sglang-milesfrom

Shi-Dong:shi/lora-release-on-abort

Commits on Jul 22, 2026

(cherry-pick) feat: add native gRPC sidecar module launcher (#31076) (#32074)

ishandhanani
and
Connor Carpenter
authored
49e709e

Commits on Jul 23, 2026

[Cherry-pick to release/v0.5.16] Fix nvfp4 online scale with pcg (#32246) (#32259)

Kangyan-Zhou
and
Qiaolin-Yu
authored
6227ae8
[Cherry-pick to release/v0.5.16] [spec decoding] fix inkling multi layer mtp draft extend cuda graph (#32254) (#32260)

Kangyan-Zhou
and
Qiaolin-Yu
authored
6870be7

Commits on Jul 24, 2026

Commits on Jul 25, 2026

[1/27] [sglang-miles] True on-policy training support (FSDP2 + qwen_dense) (#18639)

yueming-yuan
and
maocheng23
committed
5338b11
[2/27] [sglang-miles] R3 (Rollout Routing Replay) DeepEP and MTP support (#18642)
yueming-yuan
committed
726efd8
[3/27] [sglang-miles] PD disaggregation for RL (#18646)
yueming-yuan
committed
23b0989
[4/27] [sglang-miles] MTP related fix (#18647)
yueming-yuan
committed
3a01329
[5/27] [sglang-miles] VLM training multimodal fallback fixes (#18781)
yueming-yuan
committed
502c4d5
[6/27] [sglang-miles] Fix pause-aware weight update deadlocks (#22754, #22623)

maocheng23
authored and
yueming-yuan
committed
9c3ea38
[7/27] [sglang-miles] R3 support on PD disaggregation mini_lb (#22916)

zyzshishui
authored and
yueming-yuan
committed
6005dcc
[8/27] [sglang-miles] Improve PD pause handling (#23672, #23887)

ByronHsu
authored and
yueming-yuan
committed
4014c32
[9/27] [sglang-miles] Add KimiK2 raw tool call id parser (#25196)

guapisolo
authored and
yueming-yuan
committed
e08222b
[10/27] Fix GLM4 MoE Lite shared expert TP flag
yueming-yuan
committed
496d820
[11/27] [sglang-miles] MoE-LoRA: kimi 2.5/2.6, GLM-5.2 FP8, cuda-graph/RL fixes (#25141, #29874, #31251)

authored and
yueming-yuan
committed
448997a
[12/27] [sglang-miles] DeepSeek V4 RL fixes (#27131, #27603, #27604, #27728)
yueming-yuan
committed
ed1b342
[13/27] [sglang-miles] load_lora_adapter_from_distributed API, with upsert (#27268, #31759, #30913)

authored and
yueming-yuan
committed
976db7a
[14/27] [sglang-miles] P2P weight update support and fixes (#21278, #22663)

authored and
yueming-yuan
committed
9e3f38a
[15/27] [sglang-miles] rollout indexer replay: raw seq-relative topk capture + meta_info threading
yueming-yuan
committed
37f5a22
[16/27] [sglang-miles] RL weight-update sessions + distributed update for spec draft worker(s) (#27749, #28575, #18565, #22663, #28001, #29675, #27750)

committed
bb17e99
[17/27] [sglang-miles] exclude shared skip-topk layer indexer weights from RL weight check (#29339)
yueming-yuan
committed
94442dc
[18/27] [sglang-miles] Fix stale _attn_sink_local cache after RL weight updates (#30421)
yueming-yuan
committed
30bce16
[19/27] [lora] Support GDN in_proj_ba adapters for Qwen3.5 (#30499)

yushengsu-thu
authored and
yueming-yuan
committed
6276616
[20/27] [RL] Add /pull_weights: engine-side pull of published weights into a host-local checkpoint (#30366, #28524)

nanjiangwill
authored and
yueming-yuan
committed
e0bd0e6
[21/27] feat(sglang-miles): Support aborting requests by rid prefix - multi-lora needs (#30912)

mathewjhan
authored and
yueming-yuan
committed
4b64264
[22/27] [sglang-miles] Fix flush_cache() no-op after pause_generation in retract (#31962)

Zhichenzzz
authored and
yueming-yuan
committed
d0c484a
[23/27] [sglang-miles] check_weights: make the wire ChecksumInfo match the merged per-role payload
yueming-yuan
committed
dd2c725
[24/27] [sglang-miles] Port RankParallelismConfig / ParallelismContext to v0.5.16's parallel state

JD-ETH
authored and
yueming-yuan
committed
7687585
[25/27] [sglang-miles] Repoint three symbols v0.5.16 renamed or moved
yueming-yuan
committed
47325f4
[26/27] [sglang-miles] Repoint two more attributes v0.5.16 relocated
yueming-yuan
committed
119689c
[27/27] [sglang-miles] Fix iter_runners reaching the draft runner on the wrong object

guapisolo
authored and
yueming-yuan
committed
a3839f7

Commits on Jul 29, 2026

fix: honor weight-check skips for quantized entries (#32809)
guapisolo
authored
5138eb2
[Fix] Make RowParallelLinear k-size tuple-aware for FP8 (#30742)

XinyuJiangCMU
authored and
yueming-yuan
committed
49a5ef8
[sglang-miles] Warn instead of silently dropping one-sided MoE expert LoRA targets (#32376)

yushengsu-thu
authored and
yueming-yuan
committed
3003d70

Commits on Jul 30, 2026

[sglang-miles] Back up the CUDA graph pool across TMS pause/resume (#32718)
zyzshishui
authored
71db101

Commits on Jul 31, 2026

[Cherry-pick of #32861] Fix Inkling tool-call parsing recovery, content handling, and streaming (#32958)

Zhichenzzz
and
ispobock
authored
3fe50ed

Commits on Aug 3, 2026

[sglang-miles] Release the LoRA reference on every request-cleanup path
Shi-Dong
committed
0b45fbc
fix(lora): make request accounting lifecycle-safe

yushengsu-thu
and
shafeeqibraheem
committed
ffc5748

Commits on Aug 4, 2026

Commits on Aug 5, 2026

[sglang-miles] Nemotron support on sglang-miles (#27110)

Zhichenzzz
and
yueming-yuan
authored
cb05a44

Commits on Aug 10, 2026

Merge sglang-miles and preserve transactional LoRA leases
yushengsu-thu
committed
18c0019