[Doc] Add release note for v0.11.0rc1#3931
Conversation
|
👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:
If CI fails, you can run linting and testing checks locally according Contributing and Testing. |
github-actions
Bot
added
the
documentation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds the release notes for v0.11.0rc1. The changes are limited to documentation. I have found a potentially misleading statement regarding a feature's default enablement status, which could cause confusion for users. Please see the specific comment for details.
| ### Highlights | ||
| - Deepseek series models work with aclgraph now. | ||
| - PrefixCache and Chunked Prefill are enabled by default. | ||
| - W4A8 quantization is supported now. |
There was a problem hiding this comment.
Maybe we can highlight our refactor efforts here? After a series of arduous efforts, it is no longer necessary to re-register language models in vLLM-Ascend now :)
There was a problem hiding this comment.
release note is for end user, the model refactor doesn't affect users. I'll mention it in others section
| v0.11.0 will be the next official release version of vLLM Ascend. We'll release it in the next few days. Any feedback is welcome to help us improve v0.11.0. | ||
|
|
||
| ### Highlights | ||
| - Deepseek series models work with aclgraph now. |
There was a problem hiding this comment.
Will this release include aclgraph support for deepseek v3.2? Maybe it's better to narrow the scope of Deepseek series model, like Deepseek V2/V3. Deepseek series model also includes DeepSeek Ocr which might cause misunderstanding.
| ### Core | ||
| - Performance of Qwen and Deepseek series models are improved. | ||
| - Mooncake store connector, Mooncake layerwise connector, CPU offload connector are supported now | ||
| - MTP > 1 is supported now. |
There was a problem hiding this comment.
Due to vLLM v0.11.0 don't support MTP > 1, we need to set other param to support this case, and this will be explained in the MTP developer guide. Should we add a redirect link here to explain this matter?
There was a problem hiding this comment.
yes, it's fine to add any link
wangxiyuan
force-pushed
the
add_releasenote_0110rc1
branch
from
a836489 to
382a074
Compare
wangxiyuan
force-pushed
the
add_releasenote_0110rc1
branch
4 times, most recently
from
d3d53e7 to
f486a59
Compare
| - W4A4 quantization is supported now. [#3427](https://github.com/vllm-project/vllm-ascend/pull/3427) | ||
|
|
||
| ### Core | ||
| - Performance of Qwen and Deepseek series models are improved. |
| - LLMDatadist KV Connector is deprecated. We'll remove it in Q1 2026. | ||
| - Refactor the linear module to support features flashcomm1 and flashcomm2 in paper [flashcomm](https://arxiv.org/pdf/2412.04964) [#3004](https://github.com/vllm-project/vllm-ascend/pull/3004) [#3334](https://github.com/vllm-project/vllm-ascend/pull/3334) | ||
|
|
||
| ### Known issue |
There was a problem hiding this comment.
Each known issue should have a related issue link
wangxiyuan
force-pushed
the
add_releasenote_0110rc1
branch
2 times, most recently
from
d53b621 to
9251180
Compare
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
wangxiyuan
force-pushed
the
add_releasenote_0110rc1
branch
from
9251180 to
2330b25
Compare
|
|
||
| - [[v0.9.1] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/2643) | ||
| - [[v0.11.0rc0] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/3222) | ||
| - [[v0.11.0rc1] FAQ & Feedback](https://github.com/vllm-project/vllm-ascend/issues/3222) |
There was a problem hiding this comment.
plz update the issue number
Add release note for v0.11.0rc1. - vLLM version: v0.11.0 - vLLM main: vllm-project/vllm@83f478b Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Signed-off-by: luolun <luolun1995@cmbchina.com>
Add release note for v0.11.0rc1. - vLLM version: v0.11.0 - vLLM main: vllm-project/vllm@83f478b Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Signed-off-by: hwhaokun <haokun0405@163.com>
Add release note for v0.11.0rc1. - vLLM version: v0.11.0 - vLLM main: vllm-project/vllm@83f478b Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Signed-off-by: nsdie <yeyifan@huawei.com>
Add release note for v0.11.0rc1. - vLLM version: v0.11.0 - vLLM main: vllm-project/vllm@83f478b Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
7 participants
Add release note for v0.11.0rc1.