### What this PR does / why we need it?
This PR refactors the Ascend attention implementation to align with
vLLM's core interfaces, simplifying the code and improving
maintainability.
### Key Changes:
* **Align with vLLM's Attention Interface**: The `forward` method
signature in `AscendAttentionBackendImpl` now matches the base
`AttentionImpl` in vLLM, removing the custom `trace_flag`.
* **Enable Opaque Attention Operator**: By adding `opaque_attention_op`
to `AscendPlatform`, we allow vLLM to wrap our attention kernel in its
standard `vllm.unified_attention_with_output` operator. This avoids the
need for a custom call path.
* **Remove Obsolete Code**:
* The custom op `vllm.unified_ascend_attention_with_output` has been
deleted as it is now redundant.
* The `trace_flag` and its associated logic were removed, reducing code
complexity.
* An outdated quantization branch within the attention implementation
was cleaned up.
* **Improve Readability**: Renamed output variables (`output` vs.
`intermediate_output`) and added comments to clarify the in-place nature
of the attention output.
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
No extra tests needed.
- vLLM version: v0.11.0rc3
- vLLM main:
17c540a993
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
### What this PR does / why we need it?
This 'test_rejection_sampler' unit test is something wrong.
> def test_sample_recovered_tokens_pytorch_autoregressive(self):
> output_token_ids = torch.empty(2, dtype=torch.int32)
> cu_num_draft_tokens = torch.tensor([1, 1])
> draft_token_ids = torch.tensor([0, 1])
len(draft_token_ids ) = 2, cu_num_draft_tokens should be
torch.tensor([1, 2]) or torch.tensor([2, 2])
I fix it and set cu_num_draft_tokens = torch.tensor([1, 2]). The methods
before and after optimization can pass.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
NA
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: lio <1983142975@qq.com>
### What this PR does / why we need it?
Fix mooncake connector. In scenarios where TP is not equal, when the
prefill TP size is less than the number of key-value heads,
_get_remote_tp_ranks_for_req will return a list of np.arrays. Performing
an operation like int in list of np.arrays will cause an error.
Converting the list of np.arrays into a single np.array resolves this
issue.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
qwen235B
P tp16, D tp1
P tp8, D tp1
P tp4, D tp1
P tp8, D tp2
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: liziyu <liziyu16@huawei.com>
Signed-off-by: underfituu <hzhucong@163.com>
Co-authored-by: underfituu <hzhucong@163.com>
### What this PR does / why we need it?
This PR introduces a new model loader called Netloader, which leverages
high-bandwidth P2P direct transfer between NPU cards to achieve weight
loading. Netloader is implemented as a plugin through the newly added
'register_model_loader' function in vLLM 0.10. It facilitates the
process of weight loading by sending weights from a pre-loaded model
(server) to an empty model of a newly started instance (client). The
server operates concurrently with normal inference tasks through
sub-threads and the 'stateless_init_torch_distributed_process_group' in
vLLM. The client initiates a transfer request after verifying that the
model and partitioning method are the same as the server's, and uses
HCCL's collective communication (send/recv) to load the weights in the
order they are stored in the model.
Application Scenarios:
1. Significantly Reduces Inference Instance Startup Time By reusing the
weights of already loaded instances and performing high-speed transfers
directly between computing cards, this method reduces model loading
latency compared to traditional remote/local pull methods.
2. Reduces Network and Storage Pressure Avoids the need to repeatedly
download weight files from remote repositories, reducing the impact on
centralized storage and network traffic, thereby enhancing overall
system stability and service quality.
3. Improves Resource Utilization and Reduces Costs Accelerating the
loading process reduces reliance on redundant computing pools, allowing
computing resources to be elastically scaled and reclaimed as needed.
4. Enhances Business Continuity and High Availability In fault recovery
scenarios, new instances can quickly take over existing services,
avoiding prolonged business interruptions and improving the system's
high availability and user experience.
### Does this PR introduce _any_ user-facing change?
Netloader utilizes the existing --load-format=netloader and
--model-loader-extra-config to be activated. The
model-loader-extra-config needs to be input as a JSON string (as it is
now)
Afterwards, you can check whether the outputs for the same sentence are
consistent when the temperature is set to 0.
Signed-off-by: destinysky <kangrui10@126.com>
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: destinysky <kangrui10@126.com>
### What this PR does / why we need it?
This PR refactors SequenceRowParallelOp forward. In order to further
expand the operator inclusion scope in dynamic judgment scenarios, this
PR customizes the entire matmul computation and communication as a
custom operator masking. With this refactor, it will support directly
writing code such as common operation fusion into the
`SequenceRowParallelOp` class's member function `matmul_and_reduce`,
without the need to register more redundant custom masking operators.
### How was this patch tested?
CI passed with existing test.
Signed-off-by: rjg-lyh <1318825571@qq.com>
### What this PR does / why we need it?
- `qkv_proj.weight` prefetching has been implemented with `Quant` op,
when `AddRmsNormQuant` is enabled (#3465) `qkv_proj.weight` prefetching
won't work
- Implement `qkv_proj.weight` prefetching with `AddRmsNormQuant`
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
Tested on `Qwen3-235B-A22B-W8A8`
<img width="1868" height="109" alt="image"
src="https://github.com/user-attachments/assets/0bc28082-0287-4d5c-b8f6-f907c3134d36"
/>
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: zhoux77899 <zhouxiang100@huawei.com>
### What this PR does / why we need it?
This PR aims to fix performance degradation when mtp>1. Since mtp>1 may
result in more tokens (i.e. larger batch size) than acl graph maximum
batch size, this will cause draft model to run in eager mode.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: Zetong Li <slippersss@126.com>
### What this PR does / why we need it?
1. Replace manual memory cleanup with passing parameter.
2. FusedMoEPrepareAndFinalizeWithMC2 inherits All2All avoid duplicated
code.
3. Fix MC2 bug introduced in
https://github.com/vllm-project/vllm-ascend/pull/3365
4. Unify aclgraph & eager in W8A8_dynamic.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
e2e & ut
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: Pr0Wh1teGivee <calvin_zhu0210@outlook.com>
### What this PR does / why we need it?
1.Add eplb ci to check the change of eplb feature.
2.Add param checking of eplb params.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
Qwen in A3.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: offline0806 <3337230449@qq.com>
Co-authored-by: offline0806 <3337230449@qq.com>
### What this PR does / why we need it?
This reverts commit
bf87606932.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
E2E vllm serving with `enable_shared_expert_dp: true` in eager mode as
before.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: linfeng-yuan <1102311262@qq.com>
### What this PR does / why we need it?
we notice that `patch_main` is never used. Usually the patch is for all
version. And if it's for specified version, we can use `vllm_version_is`
instead. So let's remove the useless sub folder in patch module to make
it clear.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
**Problem Description:**
The existing implementation for the w4a8-dynamic linear method only
supports the old quantization format from msmodelslim. When attempting
to load models quantized with the new version, vLLM encounters errors
due to mismatched tensor shapes and unprocessed quantization parameters.
Relavant issues:
- https://github.com/vllm-project/vllm-ascend/issues/3192
- https://github.com/vllm-project/vllm-ascend/issues/3152
**Proposed Changes:**
1. Add support for w4a8 dynamic(new format) in
AscendW4A8DynamicLinearMethod and TorchairAscendW4A8DynamicLinearMethod
2. Add unit tests and e2e tests for w4a8 dynamic new and old format
models
<details>
<summary><b>details</b></summary>
1. **Support for new w4a8-dynamic format:**
* Detects quantization format by reading the "version" field in
quant_description to ensure backward compatibility.
* Handles the new pre-packed weight format (`2x int4` in an `int8`),
which has a halved dimension. It tells the vLLM loader how to unpack it
using `_packed_dim` and `_packed_factor`.
* Supports the new `scale_bias` parameter, setting its shape based on
the layer type, as required by msmodelslim. For api consistency and
future use, the `layer_type` parameter was also added to other
quantization methods.
* Updates the weight processing logic: new format weights are handled
with `.view(torch.int32)` since they're pre-packed, while old ones are
processed with `npu_convert_weight_to_int4pack`.
2. **New unit and E2E tests:**
* Added unit tests that verify the logic for both the old and new
formats.
* Split the distributed E2E test to confirm that both old and new format
models work correctly.
</details>
Theoretically, these changes will provide support for all common new
version w4a8(dynamic) models from msmodelslim.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
I implement relevant unit tests and e2e tests and test the changes with
following commands:
```bash
# unit tests
python -m pytest tests/ut/quantization/test_w4a8_dynamic.py tests/ut/torchair/quantization/test_torchair_w4a8_dynamic.py -v
# e2e tests
pytest tests/e2e/singlecard/test_quantization.py -v -s
pytest tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen3_W4A8DYNAMIC_new_version -v -s
pytest tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen3_W4A8DYNAMIC_old_version -v -s
pytest tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W4A8DYNAMIC -v -s
```
I also tested Hunyuan-1.8B-Instruct quantized with the new w4a8-dynamic
format:
```
vllm serve ./models/Hunyuan-1.8B-Instruct-quantized --gpu-memory-utilization 0.96 --quantization ascend --max-model-len 9600 --seed 0 --max-num-batched-tokens 16384
```
All tests mentioned passed locally.
**NOTE: I use quantization model from my own repo in
test_offline_inference_distributed.py**. Here is the description:
[Anionex/Qwen3-1.7B-W4A8-V1](https://modelscope.cn/models/Anionex/Qwen3-1.7B-W4A8-V1/summary)
(including quantization steps).This should be replaced by a model in
vllm-ascend ci modelscope repo.
Thanks for reading!
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: Anionex <1005128408@qq.com>
### What this PR does / why we need it?
mooncake connector support external dp & update readme
### 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: liziyu <liziyu16@huawei.com>
### What this PR does / why we need it?
Remove test cases containing CPU incompatible operators
### 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: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
The `force_attention` parameter is designed for flash infer kernel
warmup, we don't actually need it on Ascend device (at least for
now).And it tends to make things more complicated. So we replace the
`force_attention` parameter with `aclgraph_runtime_mode` in the
attention metadata creation logic.
This change makes the control flow more explicit by directly using the
graph runtime mode to determine how to build attention metadata, rather
than relying on an intermediate boolean flag. This simplification
removes redundant logic and clarifies the conditions for building
attention metadata for full decode graph mode.
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
DP + `FULL_DECODE_ONLY` + online serving.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
This reverts commit 646c1db5d7.
this new ops may lead accuracy problem
### What this PR does / why we need it?
### 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
### What this PR does / why we need it?
Port #1916 and #2157 to master branch to fuse operators in deepseek moe
layers, which can reduce scheduling overhead on devices. Note that this
feature is valid only when `tp_size = 1` and
`multistream_overlap_shared_expert` is enabled with torchair graph mode.
### Does this PR introduce _any_ user-facing change?
Users can enable this feature with `--additional-config
'{"torchair_graph_config":{"enabled":true, "enable_super_kernel":true},
"multistream_overlap_shared_expert":true}'`.
### How was this patch tested?
E2E deepseek serving with 2P1D disaggregated prefill scenarios.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: linfeng-yuan <1102311262@qq.com>
### What this PR does / why we need it?
fix skip ut test and enable ut test run normally
### 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: hfadzxy <starmoon_zhang@163.com>
This PR deletes model codes of deepseek_v2 and deepseek_v3 to reuse the
model file from vLLM.
vLLM Ascend now uses custom ops register way instead of model file
hard-coding.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: whx-sjtu <2952154980@qq.com>
This PR adds support for redundant experts in the EPLB.
Key points:
- Use global_num_experts = num_experts + num_redundant_experts
consistently.
- Backward compatible when num_redundant_experts=0.
Tested
On a 16-rank setup (W8A8) with static EPLB and expert_map_path,
verifying router logits shape and successful requests.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: yechao237 <yechao20180411@gmail.com>
### What this PR does / why we need it?
Make the Full Graph mode can run with MTP.
### 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: anon189Ty <Stari_Falcon@outlook.com>
### What this PR does / why we need it?
shared expert dp for deepseek and deepseek_mtp, could be combined with
sp to improve performance.
### 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: zhaozx-cn <zhaozx2116@163.com>
Co-authored-by: realliujiaxu <realliujiaxu@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?
Adapt deepseek-v3.2 to vllm 0.11.0, removing the useless patch.
The final goal is to remove all the patches and align the code arch to
vllm, thus we need to do the following work in next prs.
TODO:
- [x] remove patch on attention spec
- [ ] refactor the kvcache creation logic
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?
1. CI passed with existing test.
2. Test pass with deepseek-v3.2-exp
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
1. Replace manual memory cleanup with passing parameter.
2. FusedMoEPrepareAndFinalizeWithMC2 inherits All2All avoid duplicated
code.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
e2e & ut
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
Signed-off-by: Pr0Wh1teGivee <calvin_zhu0210@outlook.com>
### What this PR does / why we need it?
The `row_idx` parameter is no longer used since
PR[#2689](https://github.com/vllm-project/vllm-ascend/pull/2689), so
remove it across multiple files to remove unnecessary calculations and
parameter passing.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
accuracy test passed for Qwen3 235B and DeepSeek V3 671B after this PR.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: CaranLic <740821011@qq.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?
Fix ZeroDivisionError when prefill_tp_size > num_kv_head, in this
situation, num_head_replica can be 0 and used to divide another value,
this PR restricts the minimum value of a to be 1. And this PR fix
tp_resharding README.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
By CI.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: liziyu <liziyu16@huawei.com>
Signed-off-by: nwpu-zxr <zhouxuerong2@huawei.com>
Co-authored-by: liziyu <liziyu16@huawei.com>
### What this PR does / why we need it?
- Refacotr and integrate a unified `WeightPrefetchMethod`
- Integrate `gate_up_proj.weight` in quantized Attention modules
- Prefetching these weights ahead of matmul-like operators imporves
performance by reducing L2 cache transfer latency
### Does this PR introduce _any_ user-facing change?
Add a new config in `--additional-config` for configuration:
```json
{
"weight_prefetch_config": {
"enabled": True,
"prefetch_ratio": {
"moe": {
"gate_up": 0.8
},
},
},
}
```
This feature is enabled by default, and can be disabled through this
configuration
### 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: yuzhup <15705211260@163.com>
### What this PR does / why we need it?
Currently, when executing to the Linear layer of models in vLLM-Ascend,
the weights format is ND in unquantized case and skipped ascend case.
This PR supplements the execution logic for Linear layer. We use a new
global variable: VLLM_ASCEND_ENABLE_NZ. When VLLM_ASCEND_ENABLE_NZ=1 and
CANN version is 8.3, the weights of the Linear layer will be converted
to FRACTAL_NZ, in both unquantized case and skipped ascend case. We also
use VLLM_ASCEND_ENABLE_NZ to control the existing NZ conversion, such as
w8a8-quantized case.
### Does this PR introduce _any_ user-facing change?
Add a new global variable VLLM_ASCEND_ENABLE_NZ. If you want to use NZ
format, you should set VLLM_ASCEND_ENABLE_NZ=1.
### 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: anon189Ty <Stari_Falcon@outlook.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>
This PR will accomplish the following tasks:
**optimize SP**
In the old version implementation, the first layer was all_reduce, which
used rms to split chunks. We changed it to perform reduce_scatter on the
embedding side, replace one all_reduce operation and one chunk with one
reduce_scatter operation.
**Support qwen3 next**
Since Qwen3 Next includes a linear attention module, the prefix name of
this module cannot take effect directly.
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>
### What this PR does / why we need it?
Currently, users have to set `HCCL_BUFFSIZE` to 512~1024 to perform mc2
operators (dispatch and combine) while running moe models with large
`ep_size` and `batch_size`. This environmental variable not only affects
allocated VRAM for mc2 group, but also increases VRAM allocation for dp,
tp & ep groups, leading to significant kvcache and free_memory drops.
This PR supports to automatically calculate and set `hccl_buffer_size`
for each process group **(except mc2 group)** separately when users set
`HCCL_BUFFSIZE` for mc2 group. This can significantly reduce wasted
buffer_size set for dp, tp & ep groups.
Note that current mc2 operators can only perform communication space
partitioning based on `HCCL_BUFFSIZE` configuration. Once they support
`hccl_buffer_size` configuration with `pg_options` while initializing
process group, we'll caculate the required buffer size and users would
avoid set `HCCL_BUFFSIZE` themselves.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
We performed E2E serving with deepseek_r1 initializing DP/TP/EP/MC2
process group and observed significant kv_cache and free_memory
increase!
- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0
---------
Signed-off-by: linfeng-yuan <1102311262@qq.com>
