### Motivation
Currently dynamically experts balancing would stop-the-world.
Asynchronously expert load balancing would be better without flowing
problems:
Host-bound latency:
There are many cpu operations during EPLB such as
eplb-algorithm、creating p2p ops、and log2phy expert converting would
spend long cpu time, as ~1s.
Communication latency: The transfer time would cost much in the
situation without nvlink. As the weight of an expert maybe transfer to
multiple new positions, thus N times send/recv for one expert, with
result long latency. We had tested that batch_isend_irecv cost more
100ms for 16 experts weight transmission in A2 server of ascend.
SwiftBalancer would not stop-the-world anymore, in out test on NPU 1~2ms
cost for each layer while benefit 5ms-8ms decode latency with ep_size =
64.
The following updates have been made:
1、expert distribution recording with lower cost.
2、async cpu computing for eplb algo and other python operator.
3、new eplb algo with less expert rebalancing while almost the same
effect.
### Proposed Change
We will gradually migrate the EPLB logic to the VLLM community and
implement a generalized design. Relevant RFC:
https://github.com/vllm-project/vllm/issues/22246
The overall workflow involves:
<img width="801" height="302"
alt="474430541-23b06f58-23bc-44a3-a1be-00f268aeb15c"
src="https://github.com/user-attachments/assets/1d73a459-1b23-4b0a-812a-bf0a75debfed"
/>
1. Record experts distribution during forward. We using expert_token_num
after disptach instead of topk_ids, thus we got much smaller tensor
shape to reduce cost of hbm recording and add-operator.
2. Do all-gather for experts distribution. Using all-gather instead of
all-reduce as less traffic volume.
3. Wake up eplb worker process with experts distribution when
num_iterations comes. Run eplb algorithm in eplb worker.
4. Generate p2p send/recv ops and other operator such as log2phy would
cost long cpu time.
5. Lanch ibatch_send_recv in async_stream before forward.
6. After forward, wait for the ibatch_send_recv finish, then do uapte
expert map and expert weights.
### Co-author
Co-authored-by: raindaywhu raindaywhu@raindaywhu@ 163.con
Co-authored-by: njuyuan yuanjl19@smail.nju.edu.cn
Co-authored-by: qmkakaxi wjh1594260677@qq.com
Co-authored-by: Skywalker-EP 173723846@qq.com
- vLLM version: v0.10.2
- vLLM main:
https://github.com/vllm-project/vllm/commit/567939953b7a9cb0ded6bf0bb21a76917b8fed97
---------
Signed-off-by: offline0806 <z00858301@china.huawei.com>
Co-authored-by: offline0806 <z00858301@china.huawei.com>
### What this PR does / why we need it?
Update max_tokens and prompt in qwen3 online doc
Before:
```
"'max_tokens' or 'max_completion_tokens' is too large: 4096. This model's maximum context length is 4096 tokens and your request has 18 input tokens (4096 > 4096 - 18). None"
```
After:
```
curl http://localhost:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
"model": "/root/.cache/modelscope/hub/models/Qwen-SGlang/Qwen3-Next-80B-A3B-Instruct",
"messages": [
{"role": "user", "content": "Who are you?"}
],
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"max_tokens": 32
}'
.{"id":"chatcmpl-8ddbd65c9ddc405397219a6792feb9a0","object":"chat.completion","created":1757985049,"model":"/root/.cache/modelscope/hub/models/Qwen-SGlang/Qwen3-Next-80B-A3B-Instruct","choices":[{"index":0,"message":{"role":"assistant","content":"Hello! I am Qwen, a large-scale language model independently developed by the Tongyi Lab under Alibaba Group. I am designed to assist you in generating various","refusal":null,"annotations":null,"audio":null,"function_call":null,"tool_calls":[],"reasoning_content":null},"logprobs":null,"finish_reason":"length","stop_reason":null,"token_ids":null}],"service_tier":null,"system_fingerprint":null,"usage":{"prompt_tokens":12,"total_tokens":44,"completion_tokens":32,"prompt_tokens_details":null},"prompt_logprobs":null,"prompt_token_ids":null,"kv_transfer_params":null}
```
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- Manually test on my local env
- CI passed
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
Remove compatibility maintenance for vllm v0.10.1 and v0.10.1.1
### Does this PR introduce _any_ user-facing change?
branch main of vllm-ascend will not be compatible with vllm v0.10.1 and
v0.10.1.1
### How was this patch tested?
CI passed with existing test.
- vLLM version: v0.10.1.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/6fb27881634d89c2e70e9e5fbad1b918c0d916cf
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
The PR is for the document of the prefiller&decoder disaggregation
deloyment guide.
The scenario of the guide is:
- Use 3 nodes totally and 2 NPUs on each node
- Qwen3-30B-A3B
- 1P2D
- Expert Parallel
The deployment can be used to verify PD Disggregation / Expert Parallel
features with a slightly less resources.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
No.
- vLLM version: v0.10.1.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/e599e2c65ee32abcc986733ab0a55becea158bb4
---------
Signed-off-by: paulyu12 <507435917@qq.com>
### What this PR does / why we need it?
This PR introduces Oproj matrix tensor model parallel to achieve
decreasing of memory consumption. It only support graph mode in pure DP
scenario.
In deepseek r1 w8a8 PD disagregated Decode instance, using pure DP, with
oproj_tensor_parallel_size = 8, we have 1 ms TPOT increasing, saved 5.8
GB NPU memory per RANK. We got best performance when
oproj_tensor_parallel_size=4 without TPOT increasing.
performance data:
<img width="1442" height="442" alt="image"
src="https://github.com/user-attachments/assets/83270fc5-868a-4387-b0a9-fac29b4a376d"
/>
### Does this PR introduce _any_ user-facing change?
This PR introduces one new config in `additional_config`.
| Name | Effect | Required | Type | Constraints |
| :---------------------------- |
:--------------------------------------- | :------- | :--- |
:----------------- |
| oproj_tensor_parallel_size | Split the o_proj matrix along the row
dimension (head num * head dim) into oproj_tensor_parallel_size pieces.
| No | int | default value is None, once this value is set, the feature
will be enabled, head num * head dim must be divisible by this value. |
example
`--additional_config={"oproj_tensor_parallel_size": 8}`
### How was this patch tested?
- vLLM version: v0.10.1.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/eddaafc1c77b0690194cbd1b73747d572793838c
---------
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
Co-authored-by: zzh <zzh_201018@outlook.com>
### What this PR does / why we need it?
This PR introduces LMhead tensor model parallel to achieve decreasing of
memory consumption, and TPOT performance improvement. It support both
eager mode and graph mode.
In deepseek r1 w8a8 PD disagregated Decode instance, using pure DP, with
lmhead_tensor_parallel_size = 8, we have 1 ms TPOT optimization, saved
1.48 GB NPU memory per RANK.
performance data:
<img width="1444" height="438" alt="image"
src="https://github.com/user-attachments/assets/3c5ef0d3-a7c7-46fd-9797-4de728eb0cb0"
/>
### Does this PR introduce _any_ user-facing change?
This PR introduces one new config in `additional_config`.
| Name | Effect | Required | Type | Constraints |
| :---------------------------- |
:--------------------------------------- | :------- | :--- |
:----------------- |
| lmhead_tensor_parallel_size | Split the lm_head matrix along the
column dimension (vocab_size) into lmhead_tensor_parallel_size pieces |
No | int | default value is None, once this value is set, the feature
will be enabled, vocab_size must be divisible by this value. |
example
`--additional_config={"lmhead_tensor_parallel_size": 8}`
### How was this patch tested?
- vLLM version: v0.10.1.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/de533ab2a14192e461900a4950e2b426d99a6862
---------
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
Co-authored-by: zhangzihang <zzh_201018@outlook.com>
### What this PR does / why we need it?
The branch `br_release_MindStudio_8.1.RC2_TR5_20260624` is commercial
delivery version of modelslim in Q3, and has been verified available
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.10.1.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/7d67a9d9f93f86b74066c64c373405aa088e4a16
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
This patch add the feature branch policy.
After this patch: maintainers are allowed to create a feature branch.
Feature branches are used for collaboration and must include an RFC
link, merge plan and mentor info.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/7be5d113d8784536b79f27f24cfa91958dc291b0
Signed-off-by: Yikun Jiang <yikunkero@gmail.com>
### What this PR does / why we need it?
enable_shared_pert_dp is currently on by default. This optimization is
currently only valid for deepseek series models. The default opening
affects the accuracy of the qwen series models.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
use parameter --additional_config='{"enable_shared_expert_dp": true}'
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/d983769c41db224e0897fac2e9aefc5f57ad1122
Signed-off-by: Wang Kunpeng <1289706727@qq.com>
### What this PR does / why we need it?
To help more developers quickly get started with vLLM, we need to write
clear and easy-to-understand code documentation and technical
interpretations. This will effectively lower the learning curve, attract
more excellent contributors, and collectively build a better developer
community.
Add ModelRunner_prepare_inputs doc
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
Pass CI
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/4be02a37767f05a3fd27d66435d5cebea7a9bfe8
---------
Signed-off-by: ChenTaoyu-SJTU <ctynb@qq.com>
### What this PR does / why we need it?
- update determinitic calculation
- update support device
### Does this PR introduce _any_ user-facing change?
- Users should update ray and protobuf when using ray as distributed
backend
- Users should change to use `export HCCL_DETERMINISTIC=true` when
enabling determinitic calculation
### How was this patch tested?
N/A
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ea1292ad3ee724e44b3dfec2a26778cd614729f9
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
1.Shared Expert Sharding Strategy Update: Switched from TP-aligned to
pure DP for shared experts, enabling more efficient execution.
2.O_Proj AllReduce → ReduceScatter: Reduced communication overhead by
using ReduceScatter, made possible by pure DP sharding.
3.AllGather Postponed: Delayed to after QKV down projection to reduce
synchronization impact during prefill.
### How was this patch tested?
Adding ut case in `tests/ut/attention/test_mla_v1.py`
#### How to run
use parameter `--additional_config='{"enable_shared_expert_dp": true}'`
##### a.How to run eager mode
eg:
python -m vllm.entrypoints.openai.api_server --model=/model_path
--trust-remote-code -tp 8 -dp 2 --enable_expert_parallel --port 8002
--max-model-len 5120 --max-num-batched-tokens 16384 --enforce-eager
--disable-log-requests
--additional_config='{"ascend_scheduler_config":{"enabled":true},"enable_shared_expert_dp":
true,"chunked_prefill_for_mla":true}'
##### b.How to run graph mode
eg:
python -m vllm.entrypoints.openai.api_server --model=/model_path
--trust-remote-code -tp 8 -dp 2 --enable_expert_parallel --port 8002
--max-model-len 5120 --max-num-batched-tokens 16384
--disable-log-requests
--additional_config='{"ascend_scheduler_config":{"enabled":true},"enable_shared_expert_dp":
true,"chunked_prefill_for_mla":true,"torchair_graph_config":{"enabled":true}}'
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/9edd1db02bc6dce6da503503a373657f3466a78b
---------
Signed-off-by: Wang Kunpeng <1289706727@qq.com>
Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
Co-authored-by: SlightwindSec <slightwindsec@gmail.com>
### What this PR does / why we need it?
In fact, the kimi-k2 model is similar to the deepseek model, and we only
need to make a few changes to support it. what does this pr do:
1. Add kimi-k2-w8a8 deployment doc
2. Update quantization doc
3. Upgrade torchair support list
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/9edd1db02bc6dce6da503503a373657f3466a78b
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
Fix cann related urls in installation doc.
### Does this PR introduce _any_ user-facing change?
The users install cann manually could use the correct url after this pr
### How was this patch tested?
N/A
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/5bbaf492a6238ff517249e73151ae9989f7bea9e
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
When using deepseek series models generated by the --dynamic parameter,
if torchair graph mode is enabled, we should modify the configuration
file in the CANN package to prevent incorrect inference results.
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/7728dd77bb802e1876012eb264df4d2fa2fc6f3c
---------
Signed-off-by: Wang Kunpeng <1289706727@qq.com>