Upgrade vLLM to v0.10.0 (#1927)

### What this PR does / why we need it?
- Upgrade to v0.10.0
- Drop v0.9.2 version compatibility
- Add patch for
`vllm_ascend/patch/worker/patch_common/patch_sampler_gather_logprobs.py`
as workaround of
f3a683b7c9
for v0.10.0 and also add e2e test `test_models_prompt_logprobs`
- Pin transformers<4.54.0 as workaround of
https://github.com/vllm-project/vllm-ascend/issues/2034

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
- Test locally:
`VLLM_USE_MODELSCOPE=true pytest -sv
tests/e2e/singlecard/test_offline_inference.py::test_models_prompt_logprobs`
- CI passed

- vLLM version: v0.9.2
- vLLM main:
7728dd77bb

---------

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

tests/e2e/singlecard/test_offline_inference.py

@@ -127,3 +127,19 @@ def test_models_topk() -> None:
                     enforce_eager=True,
                     gpu_memory_utilization=0.7) as vllm_model:
         vllm_model.generate(example_prompts, sampling_params)
+
+
+def test_models_prompt_logprobs() -> None:
+
+    example_prompts = [
+        "Hello, my name is",
+    ]
+
+    with VllmRunner("Qwen/Qwen2.5-0.5B-Instruct",
+                    max_model_len=8192,
+                    dtype="float16",
+                    enforce_eager=True,
+                    gpu_memory_utilization=0.7) as vllm_model:
+        vllm_model.generate_greedy_logprobs(example_prompts,
+                                            max_tokens=5,
+                                            num_logprobs=1)

tests/ut/attention/test_attention_v1.py

@@ -3,15 +3,12 @@ from unittest.mock import MagicMock, patch
 import torch
 
 from tests.ut.base import TestBase
-from vllm_ascend.attention.attention_v1 import \
-    AscendAttentionBackendImpl092  # isort: skip
 from vllm_ascend.attention.attention_v1 import (AscendAttentionBackend,
                                                 AscendAttentionBackendImpl,
                                                 AscendAttentionMetadataBuilder,
                                                 AscendAttentionState,
                                                 AscendMetadata,
                                                 CommonAttentionState)
-from vllm_ascend.utils import vllm_version_is
 
 
 class TestAscendAttentionBackend(TestBase):
@@ -20,12 +17,8 @@ class TestAscendAttentionBackend(TestBase):
         self.assertEqual(AscendAttentionBackend.get_name(), "ASCEND")
 
     def test_get_impl_cls(self):
-        if vllm_version_is("0.9.2"):
-            self.assertEqual(AscendAttentionBackend.get_impl_cls(),
-                             AscendAttentionBackendImpl092)
-        else:
-            self.assertEqual(AscendAttentionBackend.get_impl_cls(),
-                             AscendAttentionBackendImpl)
+        self.assertEqual(AscendAttentionBackend.get_impl_cls(),
+                         AscendAttentionBackendImpl)
 
     def test_get_metadata_cls(self):
         self.assertEqual(AscendAttentionBackend.get_metadata_cls(),