### What this PR does / why we need it?
The PR is the cherry-pick from v0.9.1
https://github.com/vllm-project/vllm-ascend/pull/1953
This PR introduce a new load balance proxy server example implementation
for disaggregated pd, which support simple token&kv_cache aware load
balance routing strategy for the disaggregated pd system compared with
origin round robin toy_proxy.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
tested on real workload and unittest
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ad57f23f6a528ab01066998b41796a44340fd43d
---------
Signed-off-by: ganyi <pleaplusone.gy@gmail.com>
### What this PR does / why we need it?
Fix Qwen MoE accuracy bug in DP scenario.
Now the implentment of `FusedMoE` in vLLM use `All2AllManager` to
manager different all2all algorithm branch. And the default branch use
`Multicast` in `dispatch` phase and `all_reduce` in `combine` phase,
which are not implented in vLLM-Ascend. This leading to invoking into a
default implentment in `base_communicator`, with empty `dispatch` and
`combine` operations, thus causing the accuracy issue on it.
This pr is a temporary workaround, refacting all2all in vLLM-Ascend
could be a better way.
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ad57f23f6a528ab01066998b41796a44340fd43d
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
### What this PR does / why we need it?
We notice that vllm's main branch merged the PR
https://github.com/vllm-project/vllm/pull/21072 and
https://github.com/vllm-project/vllm/pull/21473 to support ray backend
and fix some rebase bug from previous change. Those changes makes the
disaggregate pd in vllm ascend breaks in some scenario.
In this PR, we adopt those changes to make sure the
`llmdatddist_c_mgr_connector` works fine on the newest vllm main branch.
### Does this PR introduce _any_ user-facing change?
No user face change.
### How was this patch tested?
relevant ut will be added to make sure the functionality of those
changes.
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ad57f23f6a528ab01066998b41796a44340fd43d
---------
Signed-off-by: ganyi <pleaplusone.gy@gmail.com>
### What this PR does / why we need it?
For `Qwen2.5-0.5B-Instruct` model
- the model's total number of attention heads (14) must be divisible by
tensor parallel size. (4 -> 2)
- the model does not support enable-expert-parallel
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Local Test.
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ad57f23f6a528ab01066998b41796a44340fd43d
Signed-off-by: xleoken <xleoken@163.com>
### What this PR does / why we need it?
Using vllm's AudioAsset class to retrieve remote audio
files(https://vllm-public-assets.s3.us-west-2.amazonaws.com) is not
feasible in some cases; it is recommended to switch to local retrieval.
### How was this patch tested?
vllm:main
vllm:ascend:main
results:
```bash
Adding requests: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:04<00:00, 4.62s/it]
Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:03<00:00, 3.01s/it, est. speed input: 79.03 toks/s, output: 6.31 toks/s]
generated_text: The sport referenced is soccer, and the nursery rhyme is 'Hey Diddle Diddle'.
```
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/ad57f23f6a528ab01066998b41796a44340fd43d
---------
Signed-off-by: yangqinghao-cmss <yangqinghao_yewu@cmss.chinamobile.com>
### What this PR does / why we need it?
The functions KVTransferConfig.from_cli and AscendHcclConnector are
missing in the latest vLLM version. To resolve this, I propose modifying
the kv_connector to use LLMDataDistCMgrConnector, which depends on [PR
#2079](https://github.com/vllm-project/vllm-ascend/pull/2079)
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
vllm:main
vllm-ascend:mian
results:
```bash
Adding requests: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:00<00:00, 374.27it/s]
Processed prompts: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:00<00:00, 66.06it/s, est. speed input: 449.08 toks/s, output: 66.51 toks/s]
Prefill node is finished.
INFO 07-31 09:18:30 [model_runner_v1.py:2282] Graph capturing finished in 36 secs, took 0.21 GiB
INFO 07-31 09:18:30 [core.py:201] init engine (profile, create kv cache, warmup model) took 52.49 seconds
INFO 07-31 09:18:30 [factory.py:74] Creating v1 connector with name: LLMDataDistCMgrConnector and engine_id: 28c8ced8-575c-4f87-840a-48d04d0edf7e
INFO 07-31 09:18:30 [platform.py:157] PIECEWISE compilation enabled on NPU. use_inductor not supported - using only ACL Graph mode
INFO 07-31 09:18:30 [utils.py:333] Calculated maximum supported batch sizes for ACL graph: 76
INFO 07-31 09:18:30 [utils.py:359] No adjustment needed for ACL graph batch sizes: Qwen2ForCausalLM model (layers: 24) with 67 sizes
INFO 07-31 09:18:30 [llm.py:293] Supported_tasks: ['generate']
Waiting for prefill node to finish...
Adding requests: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:00<00:00, 709.70it/s]
Processed prompts: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:00<00:00, 16.23it/s, est. speed input: 109.70 toks/s, output: 260.01 toks/s]
Prompt: 'Hello, how are you today?', Generated text: " I'm a computer program, so I don't have feelings. But I can"
Prompt: 'Hi, what is your name?', Generated text: ' I am a computer programmer. I have a question about the programming language I am'
Prompt: 'Tell me a very long story.', Generated text: ' I want to read it. I want to read it. I want to read'
Prompt: 'what is your favourite book?', Generated text: " I'm sorry, but as an AI language model, I don't have personal"
Cleanup prefill resources
All process done
```
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/9cb497bfa346721aaf5e09a7f483764a1a54f8b4
Signed-off-by: yangqinghao-cmss <yangqinghao_yewu@cmss.chinamobile.com>
### What this PR does / why we need it?
Fixed 310p failure when using the sampler feature.
The root cause is: torch_npu.npu_top_k_top_p uses the operator
aclnnApplyTopKTopP, but aclnnApplyTopKTopP currently does not support
310P.
First PR that has the issue is #1308.
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/207b750e194829c4bcd4df0450f5f93d71755dae
Signed-off-by: leo-pony <nengjunma@outlook.com>
### What this PR does / why we need it?
This PR enabled pytest and yaml style accuracy test, users now can
enable accuracy test by running:
```bash
cd ~/vllm-ascend
pytest -sv ./tests/e2e/singlecard/models/test_lm_eval_correctness.py \
--config ./tests/e2e/singlecard/models/configs/Qwen3-8B-Base.yaml \
--report_output ./benchmarks/accuracy/Qwen3-8B-Base.md
pytest -sv ./tests/e2e/singlecard/models/test_lm_eval_correctness.py \
--config-list-file ./tests/e2e/singlecard/models/configs/accuracy.txt
```
Closes: https://github.com/vllm-project/vllm-ascend/issues/1970
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/2836dd73f13015ee386c544760ca0d16888203f3
---------
Signed-off-by: Icey <1790571317@qq.com>
bugfix cherry-pick from v0.9.1-dev
https://github.com/vllm-project/vllm-ascend/pull/2007
### What this PR does / why we need it?
Minimum reproducing code:
```python
# test.py
from vllm import LLM, SamplingParams
prompts = [
"Hello, my name is",
"The future of AI is",
]
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(model="Qwen2.5-VL-7B-Instruct", max_model_len=26240)
outputs = llm.generate(prompts, sampling_params)
for output in outputs:
prompt = output.prompt
generated_text = output.outputs[0].text
print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
```
```bash
export USE_OPTIMIZED_MODEL=0
python test.py
```
exception as follow:
```
[rank0]: File "/home/xxx/vllm_ascend/models/qwen2_5_vl_without_padding.py", line 84, in forward
[rank0]: q = torch_npu.npu_rotary_mul(q, cos, sin)
[rank0]: File "/home/anaconda3/envs/xxx/lib/python3.10/site-packages/torch/_ops.py", line 1116, in __call__
[rank0]: return self._op(*args, **(kwargs or {}))
[rank0]: RuntimeError: Expected all tensors to be on the same device, but found at least two devices, npu:0 and cpu! (when checking argument for argument r1 in method wrapper__npu_rotary_mul)
```
In `AscendQwen2_5_VisionAttention_Without_Padding`,
`torch_npu.npu_rotary_mul(q, cos, sin)`, `cos`/`sin` on cpu, but `q` on
npu, so there will be an error.
`qwen2_5_vl_without_padding.py` need this bugfix, because
`AscendQwen2_5_VisionTransformer_Without_Padding.rot_pos_emb` in
wen2_5_vl_without_padding.py is from vllm and `inv_freq` will create on
cpu.
https://github.com/vllm-project/vllm/blob/40d86ee412eeeca93e0c37432db6b96829cb64e2/vllm/model_executor/models/qwen2_5_vl.py#L482
```python
inv_freq = 1.0 / (theta**(torch.arange(0, dim, 2, dtype=torch.float, device='cpu') / dim))
```
`qwen2_5_vl.py` do not need, because
`AscendQwen2_5_VisionRotaryEmbedding` in qwen2_5_vl.py rewrite
`AscendQwen2_5_VisionRotaryEmbedding` and `inv_freq` will create on
device.
```python
inv_freq = 1.0 / (theta**(torch.arange(0, dim, 2, dtype=torch.float) / dim))
```
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
CI passed with new added/existing test.
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/18cc33dd603cce058867f5730919041fa3617eb7
Signed-off-by: pjgao <gaopengju3@huawei.com>
Co-authored-by: pjgao <gaopengju3@huawei.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?
Adding `W4A8_DYNAMIC` quantization support for linear.
Dense models like Qwen3 can infer with `W4A8_DYNAMIC` quantization.
### Does this PR introduce _any_ user-facing change?
None
### How was this patch tested?
Adding ut case in `tests/ut/quantization/test_w4a8_dynamic.py`
Adding e2e case in
`tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_Qwen3_W4A8DYNAMIC`
to test qwen3 w4a8_dynamic quantized model
Note the w4a8_dynamic quantized model is quantized by `msit/msmodelslim`
of commit `d0abb0a47e1f1a473b866ad41b737fbc28fb1409`
1. Generate `W4A8_DYNAMIC` quantization weights using `msmodelslim`
```shell
git clone https://gitee.com/ascend/msit.git
cd msit/msmodelslim
git checkout d0abb0a47e1f1a473b866ad41b737fbc28fb1409
bash install.sh
```
2. Serve model using `vllm`
```shell
VLLM_USE_V1=1 python -m vllm.entrypoints.openai.api_server \
--model vllm-ascend/Qwen3-8B-W4A8 \
--port 8000 \
--quantization ascend \
--tensor_parallel_size 2 \
--enforce-eager
```
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/4cd7fe6ceaf5ad7d8ac2ba5597cd964c6db7e306
---------
Signed-off-by: ZhouXiang <zhouxiang100@huawei.com>
### What this PR does / why we need it?
This PR fixes a tensor shape mismatch in `add_lora_logits`.
Previously, `lora_a_stacked` was passed as shape `[num_loras, in_dim,
rank]`, which does not match the expected einsum pattern `"bi, boi ->
bo"` used in `bgmv_shrink`.
This causes runtime errors like:
RuntimeError: einsum(): subscript i has size 3 for operand 1 which does
not broadcast with previously seen size 4

This fix transposes `lora_a_stacked` and `lora_b_stacked` to match the
expected shapes:
- `lora_a`: `[num_loras, rank, in_dim]`
- `lora_b`: `[num_loras, out_dim, rank]`
All unit tests pass after this fix.
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?
```
import torch
import pytest
from unittest.mock import patch, PropertyMock, ANY
from vllm_ascend.lora.punica_wrapper.punica_npu import PunicaWrapperNPU
@pytest.fixture
def wrapper_cpu():
cfg = {"max_num_batched_tokens": 10, "max_batches": 2, "device": "cpu"}
w = PunicaWrapperNPU(**cfg)
w.is_prefill = True
w.no_lora = False
return w
def test_add_lora_logits(wrapper_cpu):
batch_size = 2
hidden_size = 4
lora_rank = 3
vocab_size = 5
y = torch.zeros(batch_size, vocab_size)
x = torch.randn(batch_size, hidden_size)
num_loras = 1
lora_a = torch.randn(num_loras, hidden_size, lora_rank)
lora_b = torch.randn(num_loras, lora_rank, vocab_size)
with patch.object(wrapper_cpu.__class__, "sampler_indices",
new_callable=PropertyMock) as mock_idx:
mock_idx.return_value = torch.zeros(batch_size, dtype=torch.long)
wrapper_cpu.add_lora_logits(y, x, lora_a, lora_b, scale=1.0)
assert y.shape == (batch_size, vocab_size)
assert not torch.allclose(y, torch.zeros_like(y))
Signed-off-by: hongfugui <hongfugui_yewu@cmss.chinamobile.com>
### What this PR does / why we need it?
1.Fixed the issue that pyhccl e2e cannot run continuously with other
tests.
2.Cleaned up the resources occupied by the dynamic_npugraph_batchsize
e2e test.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
This is a e2e test
e2e multi-cards tests local running successfully.
- vLLM version: v0.9.2
- vLLM main:
https://github.com/vllm-project/vllm/commit/0df4d9b06b15fa39eeb2d440e7742da93afd5e6c
Signed-off-by: leo-pony <nengjunma@outlook.com>
### What this PR does / why we need it?
Add two custom kernels(bgmv_shrink and bgmv expand) to solve the
performance of LoRA
### Does this PR introduce _any_ user-facing change?
no user-facing change
### How was this patch tested?
we add Unit Test file to test the custom ascendc kernel. See
vllm-ascend/tests/e2e/singlecard/ops/test_bgmv_expand.py and
vllm-ascend/tests/e2e/singlecard/ops/test_bgmv_expand.py
Based on the actual test of the QWen2.5 7B model using vllm-ascend
version v0.9.2.rc1, the TTFT, TPOT and throughput have increased by
about 70%.
- vLLM version: v0.9.2
- vLLM main:
https://github.com/vllm-project/vllm/commit/40d86ee412eeeca93e0c37432db6b96829cb64e2
---------
Signed-off-by: taoxudonghaha <justsheldon@163.com>
### What this PR does / why we need it?
Currently our workflow run time takes about 3 hours in total, which
seriously affects the developer experience, so it is urgent to have a
optimization, after this pr, It is expected that the running time of the
full CI can be shortened to 1h40min.
- Enable linux-aarch64-a2 (64GB) to replace linux-arm64-npu (32GB)
- Change TP4 ---> TP2 * 2 max-parallel
- Move DeepSeek-V2-Lite-W8A8 to single card test
### Does this PR introduce _any_ user-facing change?
No
- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/a2480251ec92ba2a849464dde48db8a2b7f6ef81
---------
Signed-off-by: wangli <wangli858794774@gmail.com>