realliujiaxu 5d12446573 [Feat][SP] Suport SP for VL MoE models (#7044) ...

### What this PR does / why we need it?

2nd PR for https://github.com/vllm-project/vllm-ascend/issues/5712,
extend SP to VL MoE models.


### Does this PR introduce _any_ user-facing change?
remove `sp_threshold` in additional config and reuse `sp_min_token_num`
from vLLM.


### How was this patch tested?
- Model: Qwen3-VL-30B-A3B, 
- TP4 DP2
- 100 reqs
- max concurrency 1

| Seq length | Mean TTFT (ms) main | Mean TTFT (ms) this PR |
|------------|---------------------|------------------------|
| 4k         | 429.40               | 323.3                  |
| 16k        | 1297.01              | 911.74                |

- vLLM version: v0.16.0
- vLLM main:
4034c3d32e

---------

Signed-off-by: realliujiaxu <realliujiaxu@163.com>

2026-03-24 17:16:00 +08:00

..

compile

[Feat][SP] Suport SP for VL MoE models (#7044)

2026-03-24 17:16:00 +08:00

model_runner_v2

[Feature] support aclgraph for model runner v2 (#7110)

2026-03-13 09:11:46 +08:00

pooling

[Bugfix][LoRA] Fix the bug when runs Qwen3-Reranker-0.6B with LoRA. (#7156)

2026-03-15 17:55:42 +08:00

spec_decode

[feat][spec decode]Unified draft parallel (#6766)

2026-03-13 14:07:35 +08:00

__init__.py

[CI] Add unit test framework (#1201)

2025-06-16 18:32:28 +08:00

test_aclgraph_accuracy.py

[Misc] Refactor aclgraph accuracy test to use logprob-based comparison (#7455)

2026-03-23 09:08:21 +08:00

test_aclgraph_batch_invariant.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_aclgraph_mem.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_async_scheduling.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_auto_fit_max_mode_len.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_batch_invariant.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_camem.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_completion_with_prompt_embeds.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_cpu_offloading.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_eager_mode_acc.py

[Misc] Refactor aclgraph accuracy test to use logprob-based comparison (#7455)

2026-03-23 09:08:21 +08:00

test_guided_decoding.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_ilama_lora.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_llama32_lora.py

[bugfix][LoRA] Fix the lora accuracy issue introduced by the upstream vLLM changed. (#6958)

2026-03-10 10:43:18 +08:00

test_models.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_multi_instance.py

[Bugfix] Fix multi-instance serving OOM on single card (#7427)

2026-03-23 14:22:59 +08:00

test_multistream_overlap_shared_expert.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_quantization.py

[Feature][Quant] Reapply auto-detect quantization format and support remote model ID (#7111)

2026-03-13 22:53:25 +08:00

test_qwen3_multi_loras.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_sampler.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_vlm.py

[Lint]Style: Convert test/ to ruff format(Batch #5) (#6747)

2026-02-24 15:50:00 +08:00

test_xlite.py

[Misc] Refactor aclgraph accuracy test to use logprob-based comparison (#7455)

2026-03-23 09:08:21 +08:00

utils.py

[Misc] Refactor aclgraph accuracy test to use logprob-based comparison (#7455)

2026-03-23 09:08:21 +08:00