[Bug]: Fix vllm main issue (0922)

[Yikun]

Your current environment

https://github.com/vllm-project/vllm-ascend/actions/workflows/vllm_ascend_test_full_vllm_main.yaml

# cat ~/.cache/run_single.sh
set -ex
export VLLM_WORKER_MULTIPROC_METHOD=spawn
export VLLM_USE_MODELSCOPE=True
# export VLLM_LOGGING_LEVEL=ERROR

pytest -sv tests/e2e/singlecard/test_aclgraph.py
pytest -sv tests/e2e/singlecard/test_ascend_scheduler.py
pytest -sv tests/e2e/singlecard/test_camem.py
pytest -sv tests/e2e/singlecard/test_chunked.py
pytest -sv tests/e2e/singlecard/test_embedding.py
pytest -sv tests/e2e/singlecard/test_guided_decoding.py
#pytest -sv tests/e2e/singlecard/test_ilama_lora.py
pytest -sv tests/e2e/singlecard/test_profile_execute_duration.py
pytest -sv tests/e2e/singlecard/test_quantization.py
pytest -sv tests/e2e/singlecard/test_sampler.py
pytest -sv tests/e2e/singlecard/test_vlm.py

# ------------------------------------ v1 spec decode test ------------------------------------ #
pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py
pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_torchair_correctness.py
pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py

# pytest -sv tests/e2e/singlecard/ops/

🐛 Describe the bug

0. [CI] Upgrade vLLM to 20250919 (6d8246aa) and fix some broken issue #2907

• This pr bump vllm commit to vllm-project/vllm@6d8246a
• fix upstream changes [Multimodal] Remove legacy multimodal fields in favor of MultiModalFeatureSpec  vllm#24548 abort multi-modal kwargs, make vllm main and v0.10.2 both adaptable
• fix metadata_builder changes introduced by [Core/DBO][1/N] Add Dual-Batch Overlap mechanism to VLLM vllm#23693
• fix structured_outputs_config changes introduced by [Chore] Cleanup guided namespace, move to structured outputs config vllm#22772
• fix moe_config changes introduced by [Kernel] Delegate construction of FusedMoEQuantConfig to FusedMoEMethodBase subclasses vllm#22537

1. Introduced by: vllm-project/vllm@aed1687
   Resolved: [CI] Upgrade vLLM to 20250920 (c60e613) and address config break #3067

2. vllm-project/vllm@9607d5e

TypeError: AttentionGroup.__init__() missing 1 required positional argument: 'kv_cache_spec'

Resolved: #3070

3. vllm-project/vllm@2821986
   Resolved: TBD

4. vllm-project/vllm@26e673f

VLLM_VERSION=0.10.33 ~/.cache/run_single.sh
++ export VLLM_WORKER_MULTIPROC_METHOD=spawn
++ VLLM_WORKER_MULTIPROC_METHOD=spawn
++ export VLLM_USE_MODELSCOPE=True
++ VLLM_USE_MODELSCOPE=True
++ pytest -sv tests/e2e/singlecard/test_aclgraph.py
ImportError while loading conftest '/vllm-workspace/vllm-ascend/tests/e2e/conftest.py'.
tests/e2e/conftest.py:47: in <module>
    from tests.e2e.model_utils import (TokensTextLogprobs,
tests/e2e/model_utils.py:22: in <module>
    from vllm.sequence import PromptLogprobs, SampleLogprobs
E   ImportError: cannot import name 'PromptLogprobs' from 'vllm.sequence' (/vllm-workspace/vllm/vllm/sequence.py)

diff --git a/tests/e2e/model_utils.py b/tests/e2e/model_utils.py
index 1a3ea5ba0..dec88cd0b 100644
--- a/tests/e2e/model_utils.py
+++ b/tests/e2e/model_utils.py
@@ -19,7 +19,11 @@

 from typing import Dict, List, Optional, Sequence, Tuple, Union

-from vllm.sequence import PromptLogprobs, SampleLogprobs
+from vllm_ascend.utils import vllm_version_is
+if vllm_version_is("0.10.2"):
+    from vllm.sequence import PromptLogprobs, SampleLogprobs
+else:
+    from vllm.logprobs import PromptLogprobs, SampleLogprobs

 TokensText = Tuple[List[int], str]

5. AttributeError: 'SchedulerConfig' object has no attribute 'delay_factor'
   vllm-project/vllm@0ff8ebb

[MengqingCao]

working on UniformTypeKVCacheSpecs in #3070