This PR fixes the numerical error in moe_load accumulation under ACL
graph mode on NPU: using += for NPU tensors in graph mode does not throw
errors but leads to incorrect values, so we replace it with the in-place
add_() method to ensure accurate calculation.
Signed-off-by: Mercykid-bash <ruanche0218@gmail.com>
- Fixed the computing of final hidden_states when enabling pipeline
parallel and prefill context parallel at the same time. Only in the last
PP rank, hidden_states are required and have right tensor type.
- Fixed the shape of intermediate_tensors in the dummy_run when enabling
pipeline parallel and flashcomm1. The intermediate_tensors should be
divided by tp_size. Otherwise, the moe will raise issues.
- Fixed the shape of self.intermediate_tensors for sufficient slice
space
- vLLM version: release/v0.13.0
- vLLM main:
81786c8774
---------
Signed-off-by: Jingchun Gao <gaojingchun1@huawei.com>
### What this PR does / why we need it?
For the proxy, we should remove instances when the proxy are not
processing requests.
But sometimes, We need to **isolate** some faulty nodes when a large
number of **requests** are coming in.
So we support to **isolate** faulty nodes by **lowering their priority**
and **deleted** them when the proxy does not process requests.
### Does this PR introduce _any_ user-facing change?
For
`examples/disaggregated_prefill_v1/load_balance_proxy_server_example.py`,
when using `/instances/remove` API to delete the node from the proxy
server:
```txt
curl -X POST http://localhost:9000/instances/remove \
-H "Content-Type: application/json" \
-d '{
"type": "decode",
"instances": "127.0.0.1:8201"
}'
```
There are 2 situations:
* 【New】When the proxy is processing requests, isolate the nodes and
remove them when the proxy is free.
```txt
{"message": "Instances ['127.0.0.1:8201'] are isolated and waiting to be removed.", "current_prefill_instances": ['127.0.0.1:8100', '127.0.0.1:8101'], "current_decode_instances": ['127.0.0.1:8200', '127.0.0.1:8201']}
```
* When the proxy is free, remove the nodes directly.
```txt
{"message": "remove decode instances: ['127.0.0.1:8201'].", "current_prefill_instances": ['127.0.0.1:8100', '127.0.0.1:8101'], "current_decode_instances": ['127.0.0.1:8200']}
```
### How was this patch tested?
- vLLM version: v0.13.0
- vLLM main:
11b6af5280
Signed-off-by: yuxinshan <syx_ctyg@126.com>
### What this PR does / why we need it?
Currently, the nightly image is built at 20 PM and 23 PM UTC+8. Due to
some timeliness requirements, we need to add a new trigger method for
nightly image builds.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.14.1
- vLLM main:
d68209402d
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
The static kernel in torch_npu is uninstalled through Python's atexit
mechanism.
However, in vllm-ascend, when inference ends or the service stops, the
worker process is terminated. This way, ending the process does not
trigger the atexit mechanism, causing the static kernel not to be
unloaded.
When using the nougraph_ex backend and enabling the static kernel, we
registered a signal handler to explicitly unload the static kernel.
When there are many static kernels, unloading usually takes some time,
whereas vllm will directly kill the process after sending a terminate
event. Therefore, we choose to handle it by starting a new process.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.13.0
- vLLM main:
d68209402d
---------
Signed-off-by: chencangtao <chencangtao@huawei.com>
Co-authored-by: chencangtao <chencangtao@huawei.com>
### What this PR does / why we need it?
310P support guides updates, as currently has supported in main branch.
---------
Signed-off-by: leo-pony <nengjunma@outlook.com>
### What this PR does / why we need it?
To support elastic scaling when using mooncake connector, we should
support to **configure different tp sizes for different nodes**.
As a result, we transfer the prefill node information, such as tp size,
through **the request's kv_transfer_params**.
The decode nodes **get the prefill tp size** through the request's
kv_transfer_params, instead of getting it from the configuration of the
mooncake connector .
- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef
Signed-off-by: yuxinshan <syx_ctyg@126.com>
Signed-off-by: CalvinXKY <kyxiezju@163.com>
### What this PR does / why we need it?
After removing codepsell a while, we discovered that typo had a problem
correctly recognizing certain misspelled words, so I suggested adding it
back.
- vLLM version: v0.14.1
- vLLM main:
d68209402d
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
Align with upstream vLLM. This PR will help downstream vLLM-Omni reduce
the cost for maintaining the _prepare_inputs. Besides, it helps
vLLM-Ascend code more readable. In the future, we can follow closer to
vLLM. The `preprocess` logic is same as GPUModelRunner. We don't need to
maintain it anymore.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI.
- vLLM version: v0.14.0
- vLLM main:
d68209402d
---------
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
### What this PR does / why we need it?
Fix:
```
DeprecationWarning: max_tokens is deprecated in favor of the max_completion_tokens field.
```
- vLLM version: v0.14.1
- vLLM main:
d68209402d
Signed-off-by: shen-shanshan <467638484@qq.com>
### What this PR does / why we need it?
Currently, we pad the last dim of qkv to 128 before flash attention (in
`AscendMMEncoderAttention`) to get better performance on Ascend NPU.
However, the qkv padding is executed serially, which may lead to more
overhead when launching `aclnnConstantPadNd` (launch 3 times).
Since the three operations are mutually independent, we stack qkv first
and then pad them in one kernel launch. With this optimization, **TTFT**
has been reduced by **3.15%**, **peak throughput** has been increased by
**4.20%**.
---
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Launch the server:
```bash
vllm serve /root/.cache/modelscope/hub/models/Qwen/Qwen3-VL-8B-Instruct \
--dtype bfloat16 \
--limit-mm-per-prompt '{"image": 1}' \
--max-model-len 16384 \
--max-num-batched-tokens 16384
```
Run benchmark:
```bash
vllm bench serve \
--model /root/.cache/modelscope/hub/models/Qwen/Qwen3-VL-8B-Instruct \
--backend openai-chat \
--endpoint /v1/chat/completions \
--dataset-name hf \
--hf-split train \
--dataset-path lmarena-ai/vision-arena-bench-v0.1 \
--num-prompts 1000 \
--no-stream
```
Before this PR:
```
============ Serving Benchmark Result ============
Successful requests: 1000
Failed requests: 0
Benchmark duration (s): 122.33
Total input tokens: 66638
Total generated tokens: 122845
Request throughput (req/s): 8.17
Output token throughput (tok/s): 1004.18
Peak output token throughput (tok/s): 3073.00
Peak concurrent requests: 1000.00
Total token throughput (tok/s): 1548.90
---------------Time to First Token----------------
Mean TTFT (ms): 51757.16
Median TTFT (ms): 44853.42
P99 TTFT (ms): 110700.14
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 226.06
Median TPOT (ms): 206.85
P99 TPOT (ms): 935.31
---------------Inter-token Latency----------------
Mean ITL (ms): 208.82
Median ITL (ms): 96.37
P99 ITL (ms): 2183.13
==================================================
```
After this PR:
```
============ Serving Benchmark Result ============
Successful requests: 1000
Failed requests: 0
Benchmark duration (s): 121.47
Total input tokens: 66638
Total generated tokens: 122860
Request throughput (req/s): 8.23
Output token throughput (tok/s): 1011.47
Peak output token throughput (tok/s): 3202.00
Peak concurrent requests: 1000.00
Total token throughput (tok/s): 1560.08
---------------Time to First Token----------------
Mean TTFT (ms): 50125.08
Median TTFT (ms): 46270.85
P99 TTFT (ms): 108107.12
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 227.11
Median TPOT (ms): 205.13
P99 TPOT (ms): 816.08
---------------Inter-token Latency----------------
Mean ITL (ms): 204.60
Median ITL (ms): 92.66
P99 ITL (ms): 2219.02
==================================================
```
- vLLM version: v0.14.0
- vLLM main:
d68209402d
Signed-off-by: shen-shanshan <467638484@qq.com>
### What this PR does / why we need it?
Implement `apply_top_k_top_p` via ascendC to eliminate the constraint of
k [1,1024]. It enables high performance TopKTopP calculation and avoid
D2H synchronization introduced by k validation.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
E2E serving with `k=4096` and `p=0.95`
- vLLM version: v0.13.0
- vLLM main:
d68209402d
---------
Signed-off-by: linfeng-yuan <1102311262@qq.com>
Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
Co-authored-by: SlightwindSec <slightwindsec@gmail.com>
### What this PR does / why we need it?
This patch purpose to add the `update_max_model_len` interface.
- vLLM version: v0.14.0
- vLLM main:
d68209402d
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
This patch add new runner labels for the HK region, and e2e single-card
testing has been migrated to this runner.
- vLLM version: release/v0.13.0
- vLLM main:
bc0a5a0c08
---------
Signed-off-by: wangli <wangli858794774@gmail.com>
### What this PR does / why we need it?
In vLLM-Omni, there exists the empty `ModelConfig`. We need to add a
check before accessing the sub-field of model_config.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Will checked by CI.
- vLLM version: v0.14.0
- vLLM main:
d68209402d
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
This reverts commit 8966a99710.
It breaks the test
`tests/e2e/singlecard/spec_decode/test_mtp_eagle_correctness.py::test_deepseek_mtp_correctness[True-FULL_DECODE_ONLY-2-wemaster/deepseek_mtp_main_random_bf16]`
- vLLM version: v0.14.0
- vLLM main:
d68209402d
### What this PR does / why we need it?
Since CI has integrated Triton, `fuse_qknorm_rope` is enabled by
default.
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?
CI passed with new added/existing test.
- vLLM version: v0.14.0
- vLLM main:
d68209402d
---------
Signed-off-by: wxsIcey <1790571317@qq.com>
### What this PR does / why we need it?
**Scope of Changes**:
| File Path |
| :--- |
| `vllm_ascend/distributed/kv_transfer/__init__.py` |
| `vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_connector.py` |
|
`vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_layerwise_connector.py`
|
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.14.0
- vLLM main:
d68209402d
Signed-off-by: MrZ20 <2609716663@qq.com>
### What this PR does / why we need it?
* Refactor the LayerNorm and activation operator classes to decouple the
310P device implementation from the main branch.
* Refactor `mm_encoder_attention` on 310P to use the
`torch_npu._npu_flash_attention_unpad` operator.
* Refactor the QKV inputs in the prefill stage of `attention_v1` on 310P
so they are no longer padded to 16× alignment.
* Refactor `model_runner` on 310P to align the KV-cache initialization
logic with the mainline implementation.
### Does this PR introduce _any_ user-facing change?
NO
### How was this patch tested?
use the e2e tests.
- vLLM version: v0.13.0
- vLLM main:
d68209402d
---------
Signed-off-by: Tflowers-0129 <2906339855@qq.com>
### What this PR does / why we need it?
#5790 changes default addrmsnormBias operator if custom ops is enabled.
This PR modifies AddRmsNormQuant pass to align with addrmsnormBias.
---------
Signed-off-by: Angazenn <supperccell@163.com>
### What this PR does / why we need it?
**Refactor: Unify full-graph parameter update logic**
This PR consolidates the scattered full-graph parameter update logic
into a unified approach, improving code architecture and eliminating
duplication.
**Key improvements:**
1. **Unified interface**
- Create `update_full_graph_params` as the single entry point for all
full-graph updates
- Replace multiple scattered update calls with one unified function
- Remove ~50 lines of duplicated if-else logic across
`model_runner_v1.py` and `eagle_proposer.py`
2. **Better architecture**
- Move update logic to respective Backend classes
(`AscendAttentionBackend`, `AscendMLABackend`)
- Each Backend manages its own parameter update logic internally
- Simplify caller code to just dispatch to the appropriate Backend
3. **Cleaner parameter handling**
- Remove unnecessary `pcp_size` and `dcp_size` parameter passing
- Get parallel configuration directly from distributed groups
- Consistent with how other parts of the codebase obtain these values
**Why we need it:**
- **Maintainability**: Future changes only need to be made in one place
per Backend
- **Code quality**: Follows DRY principle and Single Responsibility
Principle
- **Readability**: Cleaner, more intuitive code structure
### Does this PR introduce _any_ user-facing change?
**No.** This is a pure refactoring with no functional changes - same
behavior, cleaner code.
### How was this patch tested?
- All existing unit tests pass with updated mocks
- No new tests needed (pure refactoring, no behavior changes)
- CI validates correctness
---
- vLLM version: v0.13.0
Signed-off-by: lico67373 <918688502@qq.com>
Co-authored-by: drslark <slarksblood@qq.com>
Co-authored-by: weijinqian0 <1184188277@qq.com>
### What this PR does / why we need it?
README.md: Improved English grammar and integrated the DeepWiki badge
for Ask AI
### Does this PR introduce _any_ user-facing change?
None
### How was this patch tested?
None
- vLLM version: v0.14.0
- vLLM main:
d68209402d
---------
Signed-off-by: fyfugoyfa <zenghaolong@huawei.com>
Signed-off-by: Mitchell-xiyunfeng <3617237115@qq.com>
Co-authored-by: fyfugoyfa <zenghaolong@huawei.com>
### What this PR does / why we need it?
Remove restrictions on mooncake for IPv6
Dependencies: cann8.5、mooncake v0.3.8.post1
- vLLM version: v0.13.0
- vLLM main:
2c24bc6996
---------
Signed-off-by: liziyu <liziyu16@huawei.com>
### What this PR does / why we need it?
This PR：
1. Enhances the logic of `_skip_all_reduce_across_dp_group` to skip all
cpu dp allreduce for dense models. This is also for purpose 2.
2. Adds `_skip_all_reduce_across_dp_group` into eagle_proposer. Now
models like Qwen3-235b supports eagle3 spec decode. A typical setting
for these moe models on pd disaggregation often introduce `dp_size > 1`.
This requires `set_forward_context` to call a cpu dp allreduce to
retrieve `num_tokens_across_dp` on all cases. Skipping this allreduce
greatly improves performance.
- vLLM version: v0.14.0
- vLLM main:
d68209402d
---------
Signed-off-by: Angazenn <supperccell@163.com>
### What this PR does / why we need it?
[Doc] Update DeepSeek-V3.2 tutorail, add single-node and multi-node
deployment
- vLLM version: v0.14.0
- vLLM main:
d68209402d
Signed-off-by: menogrey <1299267905@qq.com>
### What this PR does / why we need it?
Add the setting description of cudagraph_capture_sizes, guide users to
avoid the common mistakes frequently made when using the EAGLE overlay
fullgraph.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
No need for testing
- vLLM version: v0.13.0
- vLLM main:
8be6432bda
---------
Signed-off-by: zhaomingyu <zhaomingyu13@h-partners.com>
Signed-off-by: zhaomingyu13 <zhaomingyu13@h-partners.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
### What this PR does / why we need it?
ucm_connector add has `has_connector_metadata` interface to adapt to the
latest KV connector in vLLM.
### Does this PR introduce _any_ user-facing change?
this PR doesn't introduce _any_ user-facing change.
### How was this patch tested?
- vLLM version: v0.14.0
- vLLM main:
d68209402d
Signed-off-by: UnifiedCacheManager <unifiedcachem@163.com>
### What this PR does / why we need it?
### Does this PR introduce _any_ user-facing change?
Some synchronization logic of the fusion operator copies EP *
expertPerRank int32 values. This part of data contains synchronization
signals and data.
The 512B DataBlock of Ascend A3 writes all data in the same block
atomically to the HBM.
For the DeepSeek model, when expertPerRank per device is 16, the 512B
alignment is met in both 16-device single-node and 32-device two-node
scenarios. Therefore, we check the first position of each 512B data. If
the value is not 0, it indicates that the current 512B data has been
sent.
However, for other cases where expertPerRank per device is not 16, EP *
expertPerRank does not meet the 512B alignment. If the above logic is
used for checking, there will be problems.
Therefore, here we will pad the EP * expertPerRank data length to the
length aligned to 512B.
### How was this patch tested?
- vLLM version: v0.13.0
- vLLM main:
d68209402d
---------
Signed-off-by: lhchg <lhao_cheng@163.com>
Co-authored-by: lihaocheng <lihaosheng1@h-partners.com>
### What this PR does / why we need it?
If `sp` is enabled and `tp_size` >= 16,
`torch_npu.npu_mm_reduce_scatter_base` will raises a exception.
After consulting with the operator developer, we learned that the
operator does not work when `tp` = 16.
So, we disable the operator when `tp` = 16.
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested
We started a server with `sp` enabled and `tp` = 16.
It started successfully.
```text
[0;36m(APIServer pid=1855938)[0;0m INFO: Started server process [1855938]
[0;36m(APIServer pid=1855938)[0;0m INFO: Waiting for application startup.
[0;36m(APIServer pid=1855938)[0;0m INFO: Application startup complete.
```
- vLLM version: v0.13.0
- vLLM main:
d68209402d
Signed-off-by: drslark <slarksblood@qq.com>
### What this PR does / why we need it?
This PR is to replace addRmsNorm and Add With addRmsNormBias. This way
can lead to a more effecient result.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Full Test Pass
- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef
Signed-off-by: Chen_HaoWen <chenhaowen12@huawei.com>
Co-authored-by: Chen_HaoWen <chenhaowen12@huawei.com>
