### What this PR does / why we need it?
Support chunked prefill for Qwen3Next with PCP&DCP
- vLLM version: v0.16.0
- vLLM main:
15d76f74e2
---------
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
### What this PR does / why we need it?
There is an issue with the current default logic for MLAPO (MLA Policy
Optimization). By design, MLAPO should only be enabled by default on
Decode (D) nodes. However, in hybrid (collocated prefill and decode)
scenarios, the strategy is erroneously activated during the Prefill
stage.
This PR corrects the default behavior to ensure that MLAPO is
exclusively enabled for the Decoding phase. This prevents unexpected
policy interference during Prefill and ensures optimal performance in
hybrid deployment environments.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.14.1
- vLLM main:
dc917cceb8
---------
Signed-off-by: fems14 <1804143737@qq.com>
### What this PR does / why we need it?
1) Default enable MLAPO for deepseek MLA Attention W8A8 models on PD
disagregation D Instance, for example: DeepSeekV3-W8A8,
DeepSeek-R1-W8A8.
2) Default enable MLAPO for DeepSeek SFA Attention W8A8 models,
currently is DeepSeek-V3.2-W8A8.
### Does this PR introduce _any_ user-facing change?
Don't need use manully to VLLM_ASCEND_ENABLE_MLAPO=1, to enable MLAPO
feature for deepseek w8a8 model
The effect of enabling MLAPO SFA model deployed on a single A3 Node:
Test
with:tests/e2e/nightly/single_node/models/test_deepseek_v3_2_exp_w8a8.py
dataset: gsm8k-lite，without set MTP, FULL GRAPH, has 19% promote：
未默认开启 MLAPO 时：
├─────────────────────────┤
│ TTFT │ 14055.8836 ms │
├─────────────────────────┤
│ ITL │ 66.8171 ms. │
├─────────────────────────┤
│ Output Token Throughput │ 104.9105 token/s │
├─────────────────────────┤
默认开启 MLAPO 时：
├─────────────────────────┤
│ TTFT │ 3753.1547 ms │
├─────────────────────────┤
│ ITL. │ 61.4236 ms. │
├─────────────────────────┤
│ Output Token Throughput │ 125.2075 token/s│
├─────────────────────────┤
- vLLM version: v0.13.0
- vLLM main:
2c24bc6996
---------
Signed-off-by: leo-pony <nengjunma@outlook.com>
### What this PR does / why we need it?
In long-sequence scenarios, the chunked-prefill component may encounter
dimension misalignment issues, which previously occurred during
precision testing on the code_generate_lite dataset. This PR removes
redundant computations and instead derives the value using existing
results and straightforward calculations.
- vLLM version: v0.13.0
- vLLM main:
2c24bc6996
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
### What this PR does / why we need it?
This PR makes `AscendMLAMetadataBuilder` and `AscendSFAMetadataBuilder`
properly inherit from the base class `MLACommonMetadataBuilder` in vllm
by adding `super().__init__()` calls.
**Changes:**
- Add `super().__init__()` call in `AscendMLAMetadataBuilder.__init__()`
- Add `super().__init__()` call in `AscendSFAMetadataBuilder.__init__()`
- Extract `ascend_chunked_prefill_workspace_size()` to
`vllm_ascend/attention/utils.py` to avoid code duplication
- Override `determine_chunked_prefill_workspace_size()` to support
Ascend-specific 128k tokens workspace size (vs 64k in parent class)
- Update unit tests to mock parent class `__init__` for proper isolation
**Why we need it:**
- Follow proper Python inheritance patterns by calling
`super().__init__()`
- Reduce code duplication by reusing parent class initialization logic
- Better maintainability as parent class changes will be automatically
inherited
Part of issue #5463 item 10
### Does this PR introduce _any_ user-facing change?
No, this is an internal refactoring that does not change any user-facing
behavior.
Signed-off-by: lico67373 <918688502@qq.com>
### What this PR does / why we need it?
support dsv3.2 enable both mtp and full_decode_only
PR5626 To align with the community, the branch logic was modified.
Previously, dsv32 could not reach inside the branch, and now an
additional unpadded step is required, which causes transformations in
positions and num_input_tokens, leading to changes in the cos and sin
dimensions in sfa_v1.py. This, in turn, causes an illegal shape error
when passed to the operator.
1. The unpadded function is introduced to align with the community， and
in the community the function does not have the parameters
num_input_tokens and positions.
2. The positions are split and num_input_tokens=num_actual_tokens are
used to correspond to the function name unpad, so that the padded
positions and num_input_tokens are not output.
However, in fact, attention_v1 does not use the above two parameters.
This is done because we are concerned that some people might use these
parameters later and encounter shape mismatch issues if they are not
aware of this. Therefore, we have performed the cropping.
From the perspective of the source of acquisition, positions are not
cropped, so there is actually no need to add unpad in this case.
- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef
Signed-off-by: cookieyyds <126683903+cookieyyds@users.noreply.github.com>
### What this PR does / why we need it?
Add docstrings for Metadata and MetadataBuilder classes in the attention
module to improve code readability.
Related to #5463 (Item 11: Add some comments for CommonMetadata and
others)
**Modified files:**
- `vllm_ascend/attention/context_parallel/common_cp.py`: Added comments
for `AscendPCPMetadata`, `CPChunkedContextMetadata`,
`AscendMetadataForPrefill`, `AscendMetadataForDecode`
- `vllm_ascend/attention/utils.py`: Added comments for
`AscendPrefillContextParallelMetadata`
- `vllm_ascend/attention/mla_v1.py`: Added comments for
`ChunkedContextMetadata`, `AscendMLADecodeMetadata`
- `vllm_ascend/attention/attention_v1.py`: Added comments for
`AscendMetadata`, `AscendAttentionMetadataBuilder`
- `vllm_ascend/attention/context_parallel/attention_cp.py`: Added
comments for `AscendAttentionCPMetadataBuilder`
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Documentation only, no functional changes.
Signed-off-by: lico67373 <918688502@qq.com>
### What this PR does / why we need it?
This addresses the issue brought up by #5356 and #4963, and we believe
the unnecessary conditions are the root cause.
Change the unpad trigger to be driven by actual size mismatches
(num_reqs vs base_num_reqs or scheduled vs input token counts) rather
than specific speculative-method flags. Then remove brittle workarounds
that forced request counts and sliced query start locations.
This prevents incorrect indexing and length mismatches during
speculative decoding and makes metadata unpadding more robust across
scheduling modes.
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
Tested by existing cases.
- vLLM version: v0.13.0
- vLLM main:
8be6432bda
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
## What this PR does / why we need it?
This PR fixes the `AttentionMaskBuilder` singleton initialization issue
introduced in PR #4779 and removes the unused `pcp_prefill_mask` field.
### Background
After PR #4779 made `AttentionMaskBuilder` a singleton with `@singleton`
decorator, the class constructor now requires a `device` parameter.
However, two initialization sites were still using the old parameterless
constructor, causing failures.
### Changes
1. **Fix singleton initialization**
- Fixed `AttentionMaskBuilder()` → `AttentionMaskBuilder(self.device)`
in `AscendMLAMetadataBuilder.__init__()`
- Fixed `AttentionMaskBuilder()` → `AttentionMaskBuilder(self.device)`
in `AscendAttentionMetadataBuilder.__init__()`
2. **Remove unused field**
- Removed `pcp_prefill_mask` field from
`AscendPrefillContextParallelMetadata` (never used in codebase)
- Updated related test assertions
### Related
- Issue #5463
- PR #4779 (Unify all mask generation methods)
- PR #5389 (Make AttentionMaskBuilder singleton)
## Does this PR introduce _any_ user-facing change?
No. This is an internal refactoring.
## How was this patch tested?
- ✅ Local testing: No linter errors
- ✅ Unit tests for attention modules verified
- ⏳ CI pipeline
Signed-off-by: lico67373 <918688502@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
### What this PR does / why we need it?
This PR adds multi-stream for GQA to enable computation-communication
overlap. For chunked prefill, we reduce TTFT by approximately 4%.
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: release/v0.13.0
- vLLM main:
bc0a5a0c08
---------
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
Enable fia to support sliding window function and adapt to the Gemma3
model.
- vLLM version: release/v0.13.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: nsdie <yeyifan@huawei.com>
RFC: https://github.com/vllm-project/vllm-ascend/issues/4629
Reason：
The functions related to Cp differ significantly from those of normal
MLA-Attention, but the coupling is quite severe.
Steps：
1)Extract common code AscendMLAMetadataBuilder.build to 4 functions:
build_prefill_metadata, build_decode_metadata,build_cp_metadata,
build_chunked_metadata
todo：
1)refactor function _compute_prefill_context;
2)refactor function _mla_preprocess,_mla_decode_preprocess
3）Extract public data and processing functions from the attention_cp.py
and mla_cp.py files to the common_cp file.
vLLM version: 0.13.0rc3
vLLM main:
ad32e3e19c
- vLLM version: 0.13.0rc3
- vLLM main:
ad32e3e19c
---------
Signed-off-by: wujinyuan1 <wjy9595@qq.com>
Signed-off-by: wujinyuan1 <wujinyuan1@huawei.com>
Co-authored-by: wujinyuan1 <wjy9595@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
RFC: https://github.com/vllm-project/vllm-ascend/issues/4629
Reason:
The metadata data class contains an excessive number of variables. We
will inherit the metadata of the community and simultaneously remove
some variables that are no longer needed at present.
Todo:
1. remove attn_state partly.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>
### What this PR does / why we need it?
This PR fixes some incorrect `get_current_vllm_config` calling, which
creates empty vllm_config instead.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: Angazenn <supperccell@163.com>
RFC: https://github.com/vllm-project/vllm-ascend/issues/4629
Reason:
We distinguish the branches based on the applicable scenarios of
pagedAttention and fusedInferAttention, making the code more clear.
At the same time, it is convenient for the subsequent iterations of
sliding_window and sinks and removePA ops after FIA is ready.
Todo:
remove PA ops after FIA is ready
add slidingwindow and ops for gpt_oss
replace FIA with FIA_v2
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>
### What this PR does / why we need it?
We refactored the eagle_proposer.py to adapt the framework of eagle.py
in vllm-v0.12.0, to support the logit of padded drafter batch and
async-scheduler.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: anon189Ty <Stari_Falcon@outlook.com>
Co-authored-by: drslark <slarksblood@qq.com>
### What this PR does / why we need it?
This PR adds back pa in scenarios of small batch sizes due to
performance consideration. Will remove pa once fia performs better than
pa in all scenarios.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI passed with existing test.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: whx-sjtu <2952154980@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
### What this PR does / why we need it?
refactor npu_modelrunner， we should be close to gpu_modelrunner
### Does this PR introduce _any_ user-facing change?
NO
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: zhenwenqi2024 <zhenwenqi_2022@qq.com>
Signed-off-by: zhenwenqi2024 <155598497+zhenwenqi2024@users.noreply.github.com>
### What this PR does / why we need it?
The current library only supports the FullDecodeOnly graph mode, which
enables full graph execution during the decode. This PR extends support
to allow full graph execution in both the prefill and decode, referred
to as FULL graph mode.
- vLLM version: v0.11.0
- vLLM main:
2918c1b49c
Signed-off-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
Co-authored-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
### What this PR does / why we need it?
1、qwen GQA attention_v1 optim
2、DeepSeek MLA refactor, all gather q -> all gather kv
3、modelrunner refactor for chunk prefill, we remove some code not use
- vLLM version: v0.11.0
- vLLM main:
2918c1b49c
---------
Signed-off-by: LookAround <lixushi@huawei.com>
Signed-off-by: Delphine-Nic <tanwenqin@huawei.com>
Co-authored-by: Delphine-Nic <tanwenqin@huawei.com>
### What this PR does / why we need it?
ChunkPrefill now can support Long Sequence Feature Pcp&Dcp
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI tests passed with self-test
- vLLM version: v0.11.0
- vLLM main:
83f478bb19
---------
Signed-off-by: Apocalypse990923-qshi <qiushixu@usc.edu>
Signed-off-by: Delphine-Nic <tanwenqin@huawei.com>
Co-authored-by: Delphine-Nic <tanwenqin@huawei.com>
Co-authored-by: Delphine-Nic <3834144971@qq.com>
### What this PR does / why we need it?
1、in attention_v1 module, convert bsnd t0 tnd when pcp and dcp
2、fix tochair bug: service startup problem
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.11.0
- vLLM main:
83f478bb19
Signed-off-by: weiguihua2 <weiguihua2@huawei.com>
### What this PR does / why we need it?
Remove codes of dbo.
Currently, vLLM has supported dbo with pr:
https://github.com/vllm-project/vllm/pull/23693.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.11.0rc3
- vLLM main:
17c540a993
Signed-off-by: zzzzwwjj <1183291235@qq.com>
### What this PR does / why we need it?
This PR aims to add padding logic to seq_lens、block_tables when running
in full decode scenario. Before this PR, the number of input tokens with
padding might exceeds corresponding seq_lens. For example, when running
in full decode scenario:
```
input_ids : [1, 3, 0, 0]
seq_lens: [2, 1]
query_start_loc: [0, 1, 2]
```
Here, `input_ids` is padded by 2 tokens while
`seq_lens`/`query_start_loc` are not. The mismatch between `input_ids`
and `seq_lens`/`query_start_loc` might cause some potential bugs. This
PR would change it into :
```
input_ids : [1, 3, 0, 0]
seq_lens: [2, 1, 1, 1]
query_start_loc: [0, 1, 2, 3, 4]
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: Angazenn <supperccell@163.com>
### What this PR does / why we need it?
1.qwen3 moe uses add_rms_norm_quant op instead of 'add_rms_norm op and
quant op' during quantization scene.
2.torch_npu.add_rms_norm_quant op fixed accuracy while model weights is
quantized by anti_method m4, m4 quantization is asymmetric outlier
suppression method, it will generate none-zero norm bias,
add_rms_norm_quant op updated to add this parameter to calculate.
3. add torch-npu check
### Does this PR introduce _any_ user-facing change?
new feature works if torch_npu version >= torch_npu-2.7.1.dev20250919
### How was this patch tested?
1.no special parameters to set, no new envs to set. new feature works if
torch_npu version >= torch_npu-2.7.1.dev20250919
2.use qwen3 moe quantization model to test ,such as
Qwen3-235B-A22B-W8A8, Qwen3-30B-A3B-W8A8,
Qwen3-235B-A22B-Instruct-2507-m4 (anti_method m4)
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: h30027576 <huangdong51@huawei.com>
### What this PR does / why we need it?
This pull request integrates a new `mla_preprocess` kernel to create an
optimized path for MLA (Multi-Layer Attention) decode operations on
Ascend hardware, controlled by an environment flag. The changes include
new utility functions for weight transformation, a method to prepare
weights for the fused kernel, and logic to route decode-only batches to
this new path. My review identified a critical bug in the `transdata`
utility function where padding dimensions are swapped, which will lead
to incorrect tensor shapes and kernel failures. Additionally, I've
pointed out a high-severity maintainability issue in the
trans_rope_weight function, which modifies its input in-place, and I
have provided a pure-function alternative.
### Does this PR introduce _any_ user-facing change?
No user-facing changes by default. User can enable the `mla_preprocess`
kernel in model by enable the env-var `VLLM_ASCEND_ENABLE_MLAPO`.
### How was this patch tested?
Dedicated Ascend kernels are not covered by our CI yet, so no extra
automated tests were added. Future MLA-focused regression runs will
cover this path.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: Chen Chen <0109chenchen@gmail.com>
### What this PR does / why we need it?
when infer deepseek mtp layer with multistream_moe, we should pass a
boolean to evaluate this feature and fix bugs when we are in mtp layer
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: zouyida2052 <zouyida2002@gmail.com>
### What this PR does / why we need it?
Calculate in advance the workspace memory size needed for the
PagedAttention operator to avoid deadlocks during resource cleanup. This
PR requires torch_npu version 0920 or newer.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
Co-authored-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
### What this PR does / why we need it?
Adds support for capturing the Multi-Layer Attention (MLA) decode
operation into an ACL graph. This improves performance by compiling the
attention kernel for single-token decoding.
Key changes include:
- Implementing the graph capture logic for the MLA kernel, including
workspace management and parameter updates.
- Modifying the rotary embedding (RoPE) handling to use pre-allocated
tensors, which is a requirement for graph capture.
- Adding a `build_for_graph_capture` method to the MLA metadata builder
to create dummy metadata during the graph compilation phase.
Known issues:
- Currently, MTP is not supported in FULL_DECEDE_ONLY mode -- we're
working on a fix
- We are preparing to remove update_mla_attn_params with
auto_dispatch_capture
### Does this PR introduce _any_ user-facing change?
compilation_config={
"cudagraph_mode": "FULL_DECODE_ONLY",
},
### How was this patch tested?
- vLLM version: v0.11.0
---------
Signed-off-by: panchao-hub <315134829@qq.com>
Signed-off-by: p00465316 <panchao13@huawei.com>
Co-authored-by: p00465316 <panchao13@huawei.com>
Co-authored-by: Yizhou Liu <liu_yizhou@outlook.com>
### What this PR does / why we need it?
when mtp>1, we need refresh cos ans sin in each step.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
- vLLM version: v0.11.0
Signed-off-by: zouyida2052 <zouyida2002@gmail.com>
Note: This depends on [vLLM
#25161](https://github.com/vllm-project/vllm/pull/25161) and the
torch\_npu release from September 30.
### What this PR does / why we need it?
This pull request adds `FULL_DECODE_ONLY` mode for GQA/MHA models (MLA
models like DeepSeek V3/R1 are not included). Key improvements include:
* **Reduced dispatch latency:** By replaying the entire model execution
graph at once, we cut overhead compared with multiple smaller replays.
* **Stabilized multi-device performance:** Captureing the whole model as
one static graph also mitigates the dispatch fluctuations across
devices.
* **Stream/resource savings:** Consolidating graph captures frees up
streams, allowing more graphs to be captured.
**Known issues:**
1. `_npu_paged_attention` currently manages its own workspace in
`torch_npu`, which can deadlock when synchronizing during graph replay —
we’re working on a fix.
There may be other corner cases. This PR is the first in a planned
series; we’ll continue to iterate and address remaining issues in
follow-ups.
This is essentially a port of #1503 and #1677, but includes two major
changes:
1. Let `graph_dispatcher` decide the graph mode instead of hard-coding
it in the backend, which decouples Full Graph and Piecewise Graph and
could make it possible to remove dynamo.
2. Adapt to the new `attn_group` logic, but leave a small hack in
`update_graph_params`; multi-attention models may or may not be fully
supported yet.
### Does this PR introduce _any_ user-facing change?
```python
compilation_config={
"cudagraph_mode": "FULL_DECODE_ONLY",
},
```
### How was this patch tested?
Tests included.
- vLLM version: v0.10.2
- vLLM main:
9607d5eb44
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
### What this PR does / why we need it?
Fix some ci issue and refactor modelrunner
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?
CI passed with existing test.
- vLLM version: v0.10.0
- vLLM main:
4d9c61993a
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: MengqingCao <cmq0113@163.com>
Signed-off-by: weiguihua2 <weiguihua2@huawei.com>
Co-authored-by: wangli <wangli858794774@gmail.com>
Co-authored-by: weiguihua2 <weiguihua2@huawei.com>
