EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

add release note for 0.15.0rc1 (#6839)

Browse Source 
Add release note for 0.15.0rc1

- vLLM version: v0.15.0
- vLLM main:
83b47f67b1

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2026-02-27 11:55:55 +08:00
committed by GitHub
parent a95c0b8b82
commit 3d43ed997e
4 changed files with 96 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docs/source/community/versioning_policy.md
View File

@@ -23,6 +23,7 @@ The table below is the release compatibility matrix for vLLM Ascend release.
| vLLM Ascend | vLLM | Python | Stable CANN | PyTorch/torch_npu | Triton Ascend |
|-------------|-------------------|-----------------|-------------|---------------------------------|---------------|
| v0.15.0rc1 | v0.15.0 | >= 3.10, < 3.12 | 8.5.0 | 2.9.0 / 2.9.0 | 3.2.0 |
| v0.14.0rc1 | v0.14.1 | >= 3.10, < 3.12 | 8.5.0 | 2.9.0 / 2.9.0 | 3.2.0 |
| v0.13.0 | v0.13.0 | >= 3.10, < 3.12 | 8.5.0 | 2.9.0 / 2.8.0.post2 | 3.2.0 |
| v0.13.0rc2 | v0.13.0 | >= 3.10, < 3.12 | 8.5.0 | 2.8.0 / 2.8.0.post1 | 3.2.0 |
@@ -64,6 +65,7 @@ For main branch of vLLM Ascend, we usually make it compatible with the latest vL
| Date | Event |
|------------|-------------------------------------------|
| 2026.02.26 | Release candidates, v0.15.0rc1 |
| 2026.02.06 | v0.13.0 Final release, v0.13.0 |
| 2026.01.26 | Release candidates, v0.14.0rc1 |
| 2026.01.24 | Release candidates, v0.13.0rc2 |

8
docs/source/conf.py
View File

@@ -65,15 +65,15 @@ myst_substitutions = {
# the branch of vllm, used in vllm clone
# - main branch: 'main'
# - vX.Y.Z branch: 'vX.Y.Z'
"vllm_version": "v0.14.1",
"vllm_version": "v0.15.0",
# the branch of vllm-ascend, used in vllm-ascend clone and image tag
# - main branch: 'main'
# - vX.Y.Z branch: latest vllm-ascend release tag
"vllm_ascend_version": "v0.14.0rc1",
"vllm_ascend_version": "v0.15.0rc1",
# the newest release version of vllm-ascend and matched vLLM, used in pip install.
# This value should be updated when cut down release.
"pip_vllm_ascend_version": "0.14.0rc1",
"pip_vllm_version": "0.14.1",
"pip_vllm_ascend_version": "0.15.0rc1",
"pip_vllm_version": "0.15.0",
# CANN image tag
"cann_image_tag": "8.5.0-910b-ubuntu22.04-py3.11",
# vllm version in ci

2
docs/source/faqs.md
View File

@@ -2,7 +2,7 @@
## Version Specific FAQs
- [[v0.14.0rc1] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/6148)
- [[v0.15.0rc1] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/6838)
- [[v0.13.0] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/6583)
## General FAQs

89
docs/source/user_guide/release_notes.md
View File

@@ -1,5 +1,94 @@
# Release Notes
## v0.15.0rc1 - 2026.02.26
This is the first release candidate of v0.15.0 for vLLM Ascend. Please follow the [official doc](https://docs.vllm.ai/projects/ascend/en/latest) to get started.
### Highlights
- **NPU Graph EX (npugraph_ex) Enabled by Default**: The npugraph_ex feature is now enabled by default, providing better graph optimization with integrated inductor pass and MatmulAllReduceAddRMSNorm fusion. [#6354](https://github.com/vllm-project/vllm-ascend/pull/6354) [#6664](https://github.com/vllm-project/vllm-ascend/pull/6664) [#6006](https://github.com/vllm-project/vllm-ascend/pull/6006)
- **310P MoE and W8A8 Support**: 310P now supports MoE models, W8A8 quantization, and weightNZ feature, significantly expanding hardware capabilities. [#6530](https://github.com/vllm-project/vllm-ascend/pull/6530) [#6641](https://github.com/vllm-project/vllm-ascend/pull/6641) [#6454](https://github.com/vllm-project/vllm-ascend/pull/6454) [#6705](https://github.com/vllm-project/vllm-ascend/pull/6705)
- **Qwen3-VL-MoE EAGLE Support**: Added EAGLE speculative decoding support for Qwen3-VL-MoE model. [#6327](https://github.com/vllm-project/vllm-ascend/pull/6327)
- **Kimi-K2.5 Model Support**: Added support for Kimi-K2.5 models. [#6755](https://github.com/vllm-project/vllm-ascend/pull/6755)
### Features
- **Auto-detect Quantization Format**: Quantization format can now be auto-detected from model files. [#6645](https://github.com/vllm-project/vllm-ascend/pull/6645)
- **GPT-OSS Attention Support**: Added GPT-OSS attention implementation. [#5901](https://github.com/vllm-project/vllm-ascend/pull/5901)
- **DCP Support for SFA**: Added Decode Context Parallel (DCP) support for SFA architecture. [#6563](https://github.com/vllm-project/vllm-ascend/pull/6563)
- **Mooncake Layerwise PCP Support**: Mooncake layerwise connector now supports PCP function. [#6627](https://github.com/vllm-project/vllm-ascend/pull/6627)
- **Mooncake Connector Remote PTP Size**: Mooncake connector can now get remote PTP size. [#5822](https://github.com/vllm-project/vllm-ascend/pull/5822)
- **KV Pool Sparse Attention**: KV pool now supports sparse attention. [#6339](https://github.com/vllm-project/vllm-ascend/pull/6339)
- **Batch Invariant with AscendC**: Implemented batch invariant feature with AscendC. [#6590](https://github.com/vllm-project/vllm-ascend/pull/6590)
- **Routing Replay**: Added routing replay feature. [#6696](https://github.com/vllm-project/vllm-ascend/pull/6696)
- **Compressed Tensors MoE W4A8 Dynamic Weight**: Added support for compressed tensors moe w4a8 dynamic weight quantization. [#5889](https://github.com/vllm-project/vllm-ascend/pull/5889)
- **GLM4.7-Flash W8A8 Quantization**: Added W8A8 quantization support for GLM4.7-Flash. [#6492](https://github.com/vllm-project/vllm-ascend/pull/6492)
- **DispatchGmmCombineDecode Enhancement**: DispatchGmmCombineDecode now supports bf16/float16 gmm1/gmm2 weight and ND format weight. [#6393](https://github.com/vllm-project/vllm-ascend/pull/6393)
- **RMSNorm Dynamic Quant Fusion**: Added rmsnorm dynamic quant fusion pass. [#6274](https://github.com/vllm-project/vllm-ascend/pull/6274)
- **Worker Health Check Interface**: Added `check_health` interface for worker. [#6681](https://github.com/vllm-project/vllm-ascend/pull/6681)
### Hardware and Operator Support
- **310P Support Expansion**: Multiple improvements for 310P hardware:
- Fixed attention accuracy issue on 310P. [#6803](https://github.com/vllm-project/vllm-ascend/pull/6803)
- Added weightNZ feature for 310P with quant or unquant support. [#6705](https://github.com/vllm-project/vllm-ascend/pull/6705)
- Added addrmsnorm support for 300I DUO. [#6704](https://github.com/vllm-project/vllm-ascend/pull/6704)
- 310P now supports PrefillCacheHit state. [#6756](https://github.com/vllm-project/vllm-ascend/pull/6756)
- **ARM-only CPU Binding**: Enabled ARM-only CPU binding with NUMA-balanced A3 policy. [#6686](https://github.com/vllm-project/vllm-ascend/pull/6686)
- **Triton Rope Enhancement**: Triton rope now supports index_selecting from cos_sin_cache. [#5450](https://github.com/vllm-project/vllm-ascend/pull/5450)
- **AscendC Fused Op**: Added AscendC fused op transpose_kv_cache_by_block to speed up GQA transfer. [#6366](https://github.com/vllm-project/vllm-ascend/pull/6366)
- **Rotary_dim Parameter**: Added support for rotary_dim parameter when using partial rope in rotary_embedding. [#6581](https://github.com/vllm-project/vllm-ascend/pull/6581)
### Performance
- **Multimodal seq_lens CPU Cache**: Use `seq_lens` CPU cache to avoid frequent D2H copy for better multimodal performance. [#6448](https://github.com/vllm-project/vllm-ascend/pull/6448)
- **DispatchFFNCombine Optimization**: Optimized DispatchFFNCombine kernel performance and resolved vector error caused by unaligned UB access. [#6468](https://github.com/vllm-project/vllm-ascend/pull/6468) [#6707](https://github.com/vllm-project/vllm-ascend/pull/6707)
- **DeepSeek V3.2 KVCache Optimization**: Optimized KV cache usage for DeepSeek V3.2. [#6610](https://github.com/vllm-project/vllm-ascend/pull/6610)
- **MLA/SFA Weight Prefetch**: Refactored MLA/SFA weight prefetch to be consistent with MoE weight prefetch. [#6629](https://github.com/vllm-project/vllm-ascend/pull/6629)
- **MLP Weight Prefetch**: Refactored MLP weight prefetch to be consistent with MoE model's prefetching. [#6442](https://github.com/vllm-project/vllm-ascend/pull/6442)
- **Adaptive Block Size Selection**: Added adaptive block size selection in linear_persistent kernel. [#6537](https://github.com/vllm-project/vllm-ascend/pull/6537)
- **EPLB Memory Optimization**: Reduced memory used for heat aggregation in EPLB. [#6729](https://github.com/vllm-project/vllm-ascend/pull/6729)
- **Memory Migration and Interrupt Core Binding**: Improved binding logic with memory migration and interrupt core binding functions. [#6785](https://github.com/vllm-project/vllm-ascend/pull/6785)
- **Triton Stability**: Improved Triton stability on Ascend for large grids. [#6301](https://github.com/vllm-project/vllm-ascend/pull/6301)
### Dependencies
- **Mooncake**: Upgraded to v0.3.8.post1. [#6428](https://github.com/vllm-project/vllm-ascend/pull/6428)
### Deprecation & Breaking Changes
- **ProfileExecuteDuration**: Cleaned up and deprecated ProfileExecuteDuration feature. [#6461](https://github.com/vllm-project/vllm-ascend/pull/6461)
- **Custom rotary_embedding Operator**: Removed custom rotary_embedding operator. [#6523](https://github.com/vllm-project/vllm-ascend/pull/6523)
- **USE_OPTIMIZED_MODEL**: Cleaned up unused env `USE_OPTIMIZED_MODEL`. [#6618](https://github.com/vllm-project/vllm-ascend/pull/6618)
### Documentation
- Added AI-assisted model-adaptation workflow documentation for vllm-ascend. [#6731](https://github.com/vllm-project/vllm-ascend/pull/6731)
- Added vLLM Ascend development guidelines (AGETNS.md). [#6797](https://github.com/vllm-project/vllm-ascend/pull/6797)
- Added GLM5 tutorial documentation. [#6709](https://github.com/vllm-project/vllm-ascend/pull/6709) [#6717](https://github.com/vllm-project/vllm-ascend/pull/6717)
- Added Memcache Usage Guide. [#6476](https://github.com/vllm-project/vllm-ascend/pull/6476)
- Added request forwarding documentation. [#6780](https://github.com/vllm-project/vllm-ascend/pull/6780)
- Added Benchmark Tutorial for Suffix Speculative Decoding. [#6323](https://github.com/vllm-project/vllm-ascend/pull/6323)
- Restructured tutorial documentation. [#6501](https://github.com/vllm-project/vllm-ascend/pull/6501)
- Added npugraph_ex introduction documentation. [#6306](https://github.com/vllm-project/vllm-ascend/pull/6306)
### Others
- **MTP in PD Fullgraph**: Fixed support for ALL D-Nodes in fullgraph when running MTP in PD deployment. [#5472](https://github.com/vllm-project/vllm-ascend/pull/5472)
- **DeepSeekV3.1 Accuracy**: Fixed DeepSeekV3.1 accuracy issue. [#6805](https://github.com/vllm-project/vllm-ascend/pull/6805)
- **EAGLE Refactor**: Routed MTP to EAGLE except for PCP/DCP+MTP cases. [#6349](https://github.com/vllm-project/vllm-ascend/pull/6349)
- **Speculative Decoding Accuracy**: Fixed spec acceptance rate problem in vLLM 0.15.0. [#6606](https://github.com/vllm-project/vllm-ascend/pull/6606)
- **PCP/DCP Accuracy**: Fixed accuracy issue in PCP/DCP with speculative decoding. [#6491](https://github.com/vllm-project/vllm-ascend/pull/6491)
- **Dynamic EPLB**: Fixed ineffective dynamic EPLB bug and EPLB no longer depends on a specified model. [#6653](https://github.com/vllm-project/vllm-ascend/pull/6653) [#6528](https://github.com/vllm-project/vllm-ascend/pull/6528)
- **KV Pool Mooncake Backend**: Correctly initialized head_or_tp_rank for mooncake backend. [#6498](https://github.com/vllm-project/vllm-ascend/pull/6498)
- **Layerwise Connector Recompute Scheduler**: Layerwise connector now supports recompute scheduler. [#5900](https://github.com/vllm-project/vllm-ascend/pull/5900)
- **Memcache Pool**: Fixed service startup failure when memcache pool is enabled. [#6229](https://github.com/vllm-project/vllm-ascend/pull/6229)
- **AddRMSNormQuant**: Fixed AddRMSNormQuant not taking effect. [#6620](https://github.com/vllm-project/vllm-ascend/pull/6620)
- **Pooling Code**: Fixed pooling code issues and updated usage guide. [#6126](https://github.com/vllm-project/vllm-ascend/pull/6126)
- **Context Parallel**: Fixed and unified the PD request discrimination logic. [#5939](https://github.com/vllm-project/vllm-ascend/pull/5939)
- **npugraph_ex**: Fixed duplicate pattern issue and added extra check for allreduce rmsnorm fusion pass. [#6513](https://github.com/vllm-project/vllm-ascend/pull/6513) [#6430](https://github.com/vllm-project/vllm-ascend/pull/6430)
- **RecomputeScheduler**: Fixed incompatibility of RecomputeScheduler with vLLM v0.14.1. [#6286](https://github.com/vllm-project/vllm-ascend/pull/6286)
## v0.13.0 - 2026.02.06
This is the final release of v0.13.0 for vLLM Ascend. Please follow the [official doc](https://docs.vllm.ai/projects/ascend/en/v0.13.0/) to get started.