Integrated Repositories

NPU runners have been integrated into the following repositories:

Repo	PyPI	APT	CCache	uv	BuildKit	Squid	runs-on
vllm-project/vllm-ascend	✅	✅	-	-	✅	-	✅
vllm-project/vllm-omni	-	-	-	-	-	-	-
sgl-project/sgl-kernel-npu	❌	❌	✅	✅	-	✅	✅
sgl-project/sglang	❌	❌	✅	✅	✅	✅	-
verl-project/verl	✅	✅	❌	✅	✅	✅	-
hiyouga/LlamaFactory	✅	✅	❌	✅	-	✅	-
modelscope/ms-swift	✅	✅	❌	✅	-	✅	-
triton-lang/triton-ascend	✅	❌	✅	❌	✅	✅	✅
vllm-ascend/vllm-ascend-recipes	❌	-	❌	❌	-	✅	-
verl-project/verl-omni	✅	-	❌	❌	-	✅	-
Ascend/pytorch	❌	❌	✅	❌	-	✅	-
Ascend/sglang	✅	✅	✅	✅	✅	✅	-
alibaba/ROLL	❌	-	✅	❌	-	-	✅
areal-project/AReaL	❌	-	✅	❌	❌	✅	-
fla-org/flash-linear-attention	❌	-	❌	❌	-	✅	✅
tile-ai/tilelang-mlir-ascend	✅	✅	❌	✅	✅	✅	-
verl-project/verl-SpeCo	❌	-	✅	❌	✅	✅	-

Cache audit runs daily (incremental).

✅ = confirmed in use · ❌ = confirmed NOT in use · - = unknown

Last checked: 2026-11-04

Results sourced from https://github.com/ascend-gha-runners/docs/actions/runs/41790454135