[NPU] Sync Ascend with main and consolidate NPU adaptations - #413

Open

Meihan-chen wants to merge 73 commits into

vllm-project:ascendfrom

Meihan-chen:sync/ascend-main

Commits on Jun 14, 2026

refactor(test): flatten CPU unit tests and centralize import stubs (#238)
CalvinXKY
authored
3bc70bb
docs: align vLLM docs with current defaults (#240)
aoshen02
authored
535d5fc
refactor(weight-sync): use vLLM native /update_weights, remove worker extension (#246)

aoshen02
and
claude
authored
491665d

Commits on Jun 16, 2026

Commits on Jun 17, 2026

fix(arguments): include DP in TP auto-compute default (#252)

aoshen02
and
claude
authored
0d5bfe2
[Example] Add tau-bench multi-turn tool-use example (#142)
CalvinXKY
authored
69d5fd1

Commits on Jun 18, 2026

[codex] drop stale unit CPU CI and tests (#268)
aoshen02
authored
69cb849
docker: upgrade base to vLLM 0.23.0, remove CUDA 13 build path (#253)

aoshen02
and
claude
authored
1a13bb8

Commits on Jun 20, 2026

refactor(vllm_engine): in-process launch + dataclass field introspection (#264)

aoshen02
and
claude
authored
289ee6d

Commits on Jun 21, 2026

docs: fix Qwen3-4B download repo (#276)

aoshen02
and
LironKesem
authored
9a2427b

Commits on Jun 23, 2026

Commits on Jun 25, 2026

scripts: complete slime-exact port of most scripts except for gpt-oss 20B support (#260)

aoshen02
and
claude
authored
7198547

Commits on Jun 27, 2026

[Doc] Fix broken Qwen3-4B example link in rollout_buffer README (#295)

ajinkyajawale14499
and
ajinkya-metica
authored
429b3d2

Commits on Jun 29, 2026

sync(slime #2014..#2125): 3-way merge [WIP — 45 conflict files need review] (#286)

aoshen02
and
claude
authored
d16d1dc

Commits on Jun 30, 2026

[Example] Add MemAgent long-context RL example (mem_agent) (#291)
CalvinXKY
authored
d679c75

Commits on Jul 1, 2026

Commits on Jul 3, 2026

fix(docker): /abort_requests abort-all fix + cu13 image variant (#317)

aoshen02
and
claude
authored
8d1f4cc

Commits on Jul 6, 2026

Initial ROCm support for vime (#273)

authored
e70b319

Commits on Jul 8, 2026

[Doc] Fix stale Qwen3-30B-A3B example: restore FP8-inference section + rewrite multi-node (#322)

aoshen02
and
claude
authored
789b1bc
[Doc] Fix H800→H100 typo in Qwen3-30B-A3B example (#325)

aoshen02
and
claude
authored
2702eef

Commits on Jul 11, 2026

Commits on Jul 12, 2026

revert: sync through slime #2185 (#338)
aoshen02
authored
36a84ad

Commits on Jul 13, 2026

[bugfix] Fix distributed update weights for pipeline parallel (#329)

knlnguyen1802
and
aoshen02
authored
edeb3a1

Commits on Jul 14, 2026

Commits on Jul 15, 2026

docker: bump default CU13 image to vLLM 0.25.1 (#353)
aoshen02
authored
ee16693

Commits on Jul 16, 2026

Commits on Jul 18, 2026

Add on-policy distillation example (Qwen3-8B + Qwen3-32B vLLM teacher) (#328)
CalvinXKY
authored
5cabf1f

Commits on Jul 21, 2026

ci(rocm): AMD ROCm GPU CI on Buildkite (#356)

indianspeedster
and
chandrashekharpandey
authored
c0ed6d8

Commits on Jul 26, 2026

docs: document cache-aware vLLM router support (#371)
aoshen02
authored
1fc199d

Commits on Aug 3, 2026

feat: add vLLM encoder-prefill disaggregation (#370)

Gianthard-cyh
and
aoshen02
authored
8144096

Commits on Aug 20, 2026

sync: update from Slime and upgrade to latest vLLM nightly (#386)
aoshen02
authored
af29411

Commits on Aug 23, 2026

docker: split pull-weight patch and drop upstreamed vLLM fixes (#398)
aoshen02
authored
f1d3c6b

Commits on Aug 27, 2026

docs: surface agent workflows and fix vLLM response parsing (#401)
aoshen02
authored
8ca82e0

Commits on Aug 29, 2026

Commits on Aug 30, 2026

feat: DSpark speculative decoding draft model training for RL rollout acceleration (#397)

CalvinXKY
and
aoshen02
authored
4d25bff

Commits on Aug 31, 2026

ci(rocm): rename ROCm queue to amd_mi355_vime_rl (#405)
indianspeedster
authored
85ac426

Commits on Sep 1, 2026

refactor(npu): establish S0 platformized baseline
Meihan-chen
committed
f1483a0

Commits on Sep 2, 2026

[Sync] Update to Slime v0.3.2 and vLLM nightly (#402)
aoshen02
authored
5a273f7
sync(main): merge S1 through 69cb849b
Meihan-chen
committed
2543862

Commits on Sep 3, 2026

sync(main): merge S2 through e0ca8283
Meihan-chen
committed
31782f2
fix(npu): adapt S2 arguments and VL memory budget
Meihan-chen
committed
7d9f977

Commits on Sep 4, 2026

sync(main): merge S3-S4 through 8144096e
Meihan-chen
committed
4e90cfa
fix(npu): adapt S3-S4 to vLLM 0.25.1
Meihan-chen
committed
edea686

Commits on Sep 5, 2026

ci: allow candidate image selection (#411)
aoshen02
authored
ce92eff

Commits on Sep 7, 2026

sync(main): merge S5 through 85ac4267
Meihan-chen
committed
b879eb1
fix(npu): adapt S5 to vLLM 0.27.1
Meihan-chen
committed
f930676

Commits on Sep 8, 2026

sync(main): merge S6 through ce92eff1
Meihan-chen
committed
e5f7e24
fix(npu): adapt S6 to vLLM e6bfe03a
Meihan-chen
committed
4c182d4

Commits on Sep 9, 2026

sync(ascend): merge S7 through 745d952e
Meihan-chen
committed
9d72be5
fix(npu): adapt S7 runtime and Qwen3.5 GDN
Meihan-chen
committed
7a27808
fix(npu): isolate serving FLA libraries and force spawn
Meihan-chen
committed
81ae79d
revert(npu): defer Qwen3.5 support beyond S7
Meihan-chen
committed
46f325c
test(npu): enable torch_dist ref-load by default for Qwen3-30B
Meihan-chen
committed
e0757f9

Commits on Sep 10, 2026

style(npu): satisfy pre-commit formatting checks
Meihan-chen
committed
95f6309
merge(npu): integrate GLM-4.7-Flash native G1 support
Meihan-chen
committed
1d55086
fix(npu): enable GLM MTP draft updates and graph execution
Meihan-chen
committed
edf744e

Commits on Sep 15, 2026

merge(ascend): integrate Qwen3.5 recipe follow-up
Meihan-chen
committed
207f5af
fix(npu): align scripts and examples with native runtime
Meihan-chen
committed
affeab9
chore: remove ROCm build assets and document NPU patches
Meihan-chen
committed
22dd660