### What this PR does / why we need it?
The newest version crashes in PD separation scenarios because the
function is missing the `vllm_config` parameter.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: Jade Zheng <zheng.shoujian@outlook.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 recommended configuration in the document kv_pool.md is ascend.
Modify the default value of the protocol to ascend，Improve usability
#### 1.Configure mooncake.json
The environment variable **MOONCAKE_CONFIG_PATH** is configured to the
full path where mooncake.json is located.
```
{
"local_hostname": "xx.xx.xx.xx",
"metadata_server": "P2PHANDSHAKE",
"protocol": "ascend",
"device_name": "",
"alloc_in_same_node": true,
"master_server_address": "xx.xx.xx.xx:50088",
"global_segment_size": "1GB" (1024MB/1048576KB/1073741824B/1073741824)
}
```
**local_hostname**: Configured as the IP address of the current master
node.
**metadata_server**: Configured as **P2PHANDSHAKE**.
**protocol:** Configured for Ascend to use Mooncake's HCCL
communication.
**device_name**: ""
**alloc_in_same_node**: Indicator for preferring local buffer allocation
strategy.
**master_server_address**: Configured with the IP and port of the master
service.
**global_segment_size**: Expands the kvcache size registered by the PD
node to the master.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Mooncake does not set up a protocol to launch the pooled VLLM service;
test whether the pooling function is working.
Signed-off-by: lty <linhebiwen@gmail.com>
### What this PR does / why we need it?
vllm-ascend support Ascend950 with Qwen dense model
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: wangyao <iwangyao@outlook.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
## Description
Fix the AttributeError caused by incorrect invocation of the warm-up
function in the FlashLB algorithm:
1. **Root Cause**: The warm-up function for FlashLB is defined outside
the `PolicyFlashlb` class (not a class method), but the code incorrectly
attempted to call it via the `PolicyFlashlb` class instance.
2. **Key Fix**: Clarify the invocation rule for FlashLB: when selecting
the FlashLB algorithm, the warm-up function must be called in advance to
precompile and warm up the algorithm (invoked as a standalone function),
instead of calling it through the `PolicyFlashlb` class.
3. **Impact**: Resolve the runtime error when using FlashLB, ensure the
algorithm pre-compilation/warm-up process works as expected, and avoid
attribute missing exceptions.
Signed-off-by: Mercykid-bash <ruanche0218@gmail.com>
### What this PR does / why we need it?
This PR fixes a bug in the moe_mlp module by correcting the arguments
passed to the torch_npu.npu_dequant_swiglu_quant function.It properly
converts group_list from a cumulative sum to counts for the group_index
parameter.
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: tanqingshan (A) <50050625@china.huawei.com>
Co-authored-by: tanqingshan (A) <50050625@china.huawei.com>
Similar to #2309 , this PR introduces Embedding tensor model parallel to
achieve decreasing of memory consumption. It support both eager mode and
graph mode.
And this PR refactor module tensor parallel configurations supported in
#2309, #2167, #2120, merge all config into `finegrained_tp_config` in
`additional_config`, including:
`lmhead_tensor_parallel_size`
`oproj_tensor_parallel_size`
`embedding_tensor_parallel_size`
`mlp_tensor_parallel_size`
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
Signed-off-by: zzhxx <zhangzihang23@mails.ucas.ac.cn>
Co-authored-by: zzhx1 <zzh_201018@outlook.com>
Co-authored-by: chenxiao <Jaychou1620@Gmail.com>
Co-authored-by: zzhxx <zhangzihang23@mails.ucas.ac.cn>
Co-authored-by: Jade Zheng <zheng.shoujian@outlook.com>
In the PD separation scenario, the D node does not need to perform get
operations, and therefore does not need to create ZeroMQ (ZMQ)
communication.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
### What this PR does / why we need it?
Remove FusedMoEState which is used by torchair.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
e2e & ut
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: weichen <calvin_zhu0210@outlook.com>
### What this PR does / why we need it?
Currently, the initialization and fundamental functions of
RecomputeScheduler are broken with `vLLM v0.12.0`. This PR fixes the
conflicts of `RecomputeScheduler` and refactor its implementations by
inheriting original `Scheduler` of vLLM. Meanwhile, this PR also
supports async cheduling with recompute scheduler by implementing
`AsyncRecomputeScheduler` which is simply inherited `AsncyScheduler` of
vLLM and `RecomputeScheduler` of vLLM-Ascend with python MRO.
### Does this PR introduce _any_ user-facing change?
No. The switch naming is the same as v0.11.0 :
`recompute_scheduler_enable`
### How was this patch tested?
E2E serving with 2P1D dsv3.1 passed. The performance was the same as
original vllm scheduler with `async_scheduling` and preempted requests
in D Nodes are successfully transfered to Proxy and further to P Node.
This significantly improves the performance and robustness of PD
disaggregation deployments.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: linfeng-yuan <1102311262@qq.com>
mindie_turbo is out of data for long time. This PR remove the related register method.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Previously, if the KVCacheSendingThread couldn't create a socket because
of port conflicts or other problems, the main thread would wait
endlessly for the ready_event signal, causing the entire engine
initialization to freeze. This update fixes the issue by adding timeouts
for thread startup and handling unexpected thread exits, so the
initialization process no longer gets stuck indefinitely.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: Jade Zheng <zheng.shoujian@outlook.com>
### What this PR does / why we need it?
This PR standardizes the fusion naming, changing
`enable_quantization_fusion` to `fuse_norm_quant`, and enables e2e
testing.
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?
CI passed with new added/existing test.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: wxsIcey <1790571317@qq.com>
### What this PR does / why we need it?
The variable ’is_deepseek_v3_r1‘ now is useless in the repository, so
delete it now. And the funciton 'get_fused_moe_state' is used only for
torchair, so it need to be deleted along with torchair
### 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: hust17yixuan <303660421@qq.com>
### What this PR does / why we need it?
Current mooncake connector has following problems with PP and MTP
enabled:
1. MTP layer kv caches are not transfered, it may cause decreasing of
accept ratio: This PR add MTP layer indices for last PP stage after
calculating end_layer in transfer_kv_cache
2. While MTP enabled, PP layers divided by default may cause imbalance
between stages, we need to use `VLLM_PP_LAYER_PARTITION` environment to
make it balance by hand, but in mooncake connector kv transfer, decode
doesn't know the partition of prefill node: This PR add config
`pp_layer_partition` in `kv_connector_extra_config` to make decode node
acquire the partition information of prefill node.
### Does this PR introduce _any_ user-facing change?
When prefill using `VLLM_PP_LAYER_PARTITION` environment, add
`pp_layer_partition` in `kv_connector_extra_config` like below:
```
export VLLM_PP_LAYER_PARTITION=33,28
"kv_connector_extra_config": {
"use_ascend_direct": true,
"prefill": {
"dp_size": 1,
"tp_size": 8,
"pp_size": 2,
"pp_layer_partition": "33,28"
},
"decode": {
"dp_size": 16,
"tp_size": 1,
"pp_size": 1
}
}
```
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: lidenghui <lidenghui1110@gmail.com>
### What this PR does / why we need it?
This PR adjusts the layer prefix matching rules for tensor parallelism
(column/row parallel ops) to fit Bailing model's naming conventions
(adding "query_key_value" for column parallel and "attention.dense" for
row parallel), enabling flashcomm1 to work properly on the Bailing
model.
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: hwhaokun <haokun0405@163.com>
### What this PR does / why we need it?
This PR fix the bug in sfa-cp under multi-DP scenarios.
### Does this PR introduce _any_ user-facing change?
None
### How was this patch tested?
None
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: zzhxx <2783294813@qq.com>
Co-authored-by: clrs97 <524936896@qq.com>
The bmm_transpose operator in version 3.2 is only used in the decoding stage due to shape limitations.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: ChrisGelhLan <33011886+xlan-huawei@users.noreply.github.com>
### What this PR does / why we need it?
Support triton causal_conv1d_fn ops.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
CI passed with new added/existing test.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: QilaiZhang <245706640@qq.com>
This PR clean up useless torchair logic in model runner. The moge doc is
only for torchair, it can be removed as well.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
VLLM_ASCEND_ENABLE_TOPK_TOPP_OPTIMIZATION is enabled by default for long
time. Let's remove it now.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Update patch doc. After this PR is merged, all the new patch PR should
update this doc as well.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
This PR eliminates the simplicit HD synchronization in sfa backend, and
_build_dummy_attn_metadata and dummy_run in mtp_proposer, significantly
improving dsv3.2 performance in low-latency scenarios.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Performance improvements are observed with E2E performance serving (P:
DP4TP8EP32 D: DP8TP4EP32) with `num_speculative_tokens=3`.
DSV3.2-W8A8-EXP:
TPOT: 41.67ms -> 23.36ms
ITL: 85.93ms -> 55.96ms
DSV3.2-W8A8 (relaesed in December):
TPOT: 18.11ms
ITL: 56.13ms
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: linfeng-yuan <1102311262@qq.com>
### What this PR does / why we need it?
We introduced the npugraph_ex backend through the vllm's adaptor
dispatch mechanism to accelerate aclgraph. This solution is based on
torch.compile and uses torchair to optimize the fx.graph. The
performance gains are mainly obtained from the static kernel. We
conducted tests on Qwen3-30B and achieved over 5% performance
optimization.
### Does this PR introduce _any_ user-facing change?
Yes, we add a new switch named"enable_npugraph_ex" in additional_config,
default is False.
We also add an example to show how to register custom replacement pass
### More information about this PR
This feature depends on the release of CANN and torch_npu in Q4.
We tested it on a package that has not been publicly released yet and
verified that the functionality works.
This feature is still experimental at the moment; setting the config
true will directly raise error.
Merging into the main branch initially involves some preliminary commits
to facilitate subsequent development and testing of the feature, as well
as to avoid submitting an excessively large PR at once.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: chencangtao <chencangtao@huawei.com>
Signed-off-by: ChenCangtao <50493711+ChenCangtao@users.noreply.github.com>
Co-authored-by: chencangtao <chencangtao@huawei.com>
Co-authored-by: panchao-hub <315134829@qq.com>
Co-authored-by: wbigat <wbigat@163.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
### What this PR does / why we need it?
Support for mlapo in deepseekv3.1 w4a8, since the csrc of mlapo requires the input args `enable_inner_out` and `inner_out`, we add the dummy args here
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: GDzhu01 <809721801@qq.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
…
### What this PR does / why we need it?
When speculative decoding is enabled and temperature > 0, bonus_logits
and target_logits are sampled separately:
1. bonus_logits are sampled using a fused torch_npu.npu_top_k_top_p
operator invoked inside the main sampler,
2. while target_logits are sampled within the rejection sampler using a
less-optimized implementation composed of smaller operators.
Consequently, the cumsum operation in the top-p sampling for
target_logits becomes especially time-consuming, leading to performance
degradation.
<img width="1029" height="623" alt="image"
src="https://github.com/user-attachments/assets/1969f561-6aa5-41b3-9a87-1f64d4321cbf"
/>
Apply the fused operator to the sampling of target_logits as well to
reduce overhead
<img width="1039" height="572" alt="image"
src="https://github.com/user-attachments/assets/1e6563da-3418-405d-b657-7bbe10dd0924"
/>
### 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: funanyang <985619145@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
### What this PR does / why we need it?
The first commit support `FULL_DECODE_ONLY`:
- Update `AscendSFAMetadataBuilder` to use `num_input_tokens` for
slicing slots and positions, ensuring fixed tensor shapes.
- Implement padding logic for `query_start_loc` in `NPUModelRunner` to
support uniform decode in full graph mode, aligning with GPU runner
behavior.
- Adjust MLA cosine cache allocation to occur independently of graph
mode and switch to using device-resident sequence lengths for attention
metadata.
- Remove redundant slicing of hidden states and outputs in
`AscendSFAImpl` and optimize `sin`/`cos` cache updates.
The second commit take MTP into account:
- Update `AscendSFAMetadataBuilder` to use `num_input_tokens` for
slicing slots and positions, ensuring fixed tensor shapes.
- Implement padding logic for `query_start_loc` in `NPUModelRunner` to
support uniform decode in full graph mode, aligning with GPU runner
behavior.
- Adjust MLA cosine cache allocation to occur independently of graph
mode and switch to using device-resident sequence lengths for attention
metadata.
- Remove redundant slicing of hidden states and outputs in
`AscendSFAImpl` and optimize `sin`/`cos` cache updates.
And the rest of them are just bugfix.
### Does this PR introduce _any_ user-facing change?
None.
### How was this patch tested?
Test cases needed.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
### What this PR does / why we need it?
The logprobs_tensor was not initialized before accessing its token_id
member, leading to a crash when tokenizer.decode() is called by passing
a negative token_id
### How was this patch tested?
Constructed an inference request with two prompts and set
SamplingParams(prompt_logprobs=<non-None value>) (e.g.,
prompt_logprobs=1).
After applying the fix (proper initialization of logprobs_tensor), the
same request completed successfully without errors, and the returned
logprobs matched expected values.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: jiangweixiang <jwx02384838@antgroup.com>
Co-authored-by: jiangweixiang <jwx02384838@antgroup.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
### What this PR does / why we need it?
To support pipeline parallel with PD disaggregation, this PR support PP
in mooncake connector and fix other bugs when enable pp with other
optimization params, including following changes:
- mooncake connector support pp in prefill, we do not support decode pp
currently
- fix bugs when enable both pp and flashcomm1
- optimize ascend-scheduler to support full batch in multiple pipeline
stages, original implementation would cause all pipeline stages
batch_size total summed to max_num_seq, which makes pipeline is not
full, this optimization can make all stages running with full batch_size
= max_num_seq, the same changes will contribute to vllm scheduler too.
### Does this PR introduce _any_ user-facing change?
add `pp_size` in mooncake connector kv_connector_extra_config
```
"kv_connector_extra_config": {
"use_ascend_direct": true,
"prefill": {
"dp_size": 1,
"tp_size": 4,
"pp_size": 4
},
"decode": {
"dp_size": 16,
"tp_size": 1
}
}
```
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: chenxiao <Jaychou1620@Gmail.com>
Signed-off-by: Kurumi5210 <Jaychou1620@Gmail.com>
Signed-off-by: Kurumi5210 <jaychou1620@gmail.com>
Signed-off-by: 秋刀鱼 <jaychou1620@Gmail.com>
Co-authored-by: chenxiao <Jaychou1620@Gmail.com>
Co-authored-by: zss <zss@qq.com>
Co-authored-by: zss <3265779424@qq.com>
### What this PR does / why we need it?
Adds W4A16 quantization method for the Kimi-K2-Thinking model and
updates relevant modules to support the new quantization method.
- Implements complete W4A16 quantization method including weight
packing/unpacking, per-group quantization parameter generation,
post-processing logic and MoE method application.
- Adds parameters `use_int4_w4a16`, `w1_offset` and `w2_offset`, adjusts
`with_quant` conditional logic to support W4A16 matrix multiplication.
- Adds `packed_modules_model_mapping` for Kimi-K2-Thinking model and
processing logic for `weight_packed` field.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: zhoux77899 <zhouxiang100@huawei.com>
Signed-off-by: Ruri <33858552+zhoux77899@users.noreply.github.com>
Signed-off-by: Ruri <zhouxiang100@huawei.com>
### What this PR does / why we need it?
Support pooling models (like `bge-reranker-v2-m3`) in vllm-ascend, this
pr covered the three model types of embed (cls_token, mean_token,
lasttoken).
After this
[commit](17373dcd93),
vllm has provided support for adapting pooling models on the v1 engine.
This PR includes corresponding adaptations on the vllm-ascend side.
Fixes#1960
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: lianyibo <lianyibo1@kunlunit.com>
Signed-off-by: MengqingCao <cmq0113@163.com>
Co-authored-by: MengqingCao <cmq0113@163.com>
aclgraph is stable and fast now. Let's drop torchair graph mode now.
TODO: some logic to adapt torchair should be cleaned up as well. We'll
do it in the following PR.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
### What this PR does / why we need it?
When `hf_quant_cfg` is not None and `hf_quant_cfg.quant_method == ""`,
func `override_quantization_method` will return None and raise
ValidationError.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: zzzzwwjj <1183291235@qq.com>
### What this PR does / why we need it?
As support for the mooncake connector is now available, the llmdatadist
connector is no longer being maintained, so the llmdatadist-related
files need to be retired.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By ci
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: wangxiaoteng <wangxiaoteng@huawei.com>
Signed-off-by: liziyu <liziyu16@huawei.com>
Co-authored-by: liziyu <liziyu16@huawei.com>
### What this PR does / why we need it?
This PR is to fix a smoking test failure. Adjust mtp_proposer and
model_runner_v1 to route MTP decoding through the non‑fused MoE
implementation while keeping the overall inference flow unchanged.
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: mojave2 <chenchen145@huawei.com>
Co-authored-by: Mengqing Cao <cmq0113@163.com>
What this PR does / why we need it?
Improve usability，local_buffer_size support for units: GB, MB, KB, B,
For example, "2GB"
{
"local_hostname": "XXX.XXX.XXX.XXX",
"metadata_server": "P2PHANDSHAKE",
"protocol": "ascend",
"device_name": "",
"use_ascend_direct": true,
"master_server_address": "XXX.XXX.XXX.XXX:50088",
"global_segment_size": 60000000000,
"local_buffer_size": "2GB"
}
Does this PR introduce any user-facing change?
local_buffer_size support for units: GB, MB, KB, B
How was this patch tested?
Mooncake configures local_buffer_size as GB, MB, KB, B
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: lty <linxianchong1@huawei.com>
### What this PR does / why we need it?
bmm transpose ops can't be used in cp, so add judgement in the modeling
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: hust17yixuan <303660421@qq.com>
### What this PR does / why we need it?
Support pp for kv pool
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
---------
Signed-off-by: baxingpiaochong <771405853@qq.com>
### What this PR does / why we need it?
Fix incorrect MLAPO weight release in PD mixex scenarios.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: ZYang6263 <zy626375@gmail.com>
Co-authored-by: wangxiyuan <wangxiyuan1007@gmail.com>
### What this PR does / why we need it?
In reinforcement learning scenarios, the current inference applies a
transpose operation to the weights. For a cleaner architecture, the
weight transpose module was moved to wakeup.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c
Signed-off-by: lhp-deep <liuhaopeng1@huawei.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
