[CI] Update vllm version to 20250922(5aeb925) (#3091)

### What this PR does / why we need it?
This pr bump vllm commit hash to
5aeb925452
fix issues:  
1. https://github.com/vllm-project/vllm/pull/25345 has remove v0
metadata
2. https://github.com/vllm-project/vllm/pull/25332
3. https://github.com/vllm-project/vllm/pull/25334
4. https://github.com/vllm-project/vllm/pull/23558, note that this vllm
commit update the model register logic, which will check all the model
registered have the `vllm.model_executor.models` path , which breaks our
custom registration of the deepseek_v3 model (it doesn't exist in the
vllm model path). so I move deepseek_v3 model registy to deepseek_v2 to
solve temporary

### How was this patch tested?

- vLLM version: v0.10.2
- vLLM main:
9607d5eb44

---------

Signed-off-by: wangli <wangli858794774@gmail.com>

vllm_ascend/models/__init__.py

@@ -25,7 +25,7 @@ def register_model():
 
     ModelRegistry.register_model(
         "DeepseekV3ForCausalLM",
-        "vllm_ascend.models.deepseek_v3:CustomDeepseekV3ForCausalLM")
+        "vllm_ascend.models.deepseek_v2:CustomDeepseekV3ForCausalLM")
 
     ModelRegistry.register_model(
         "DeepSeekMTPModel",

vllm_ascend/models/deepseek_mtp.py

@@ -28,7 +28,6 @@ from vllm.config import (CacheConfig, ModelConfig, VllmConfig,
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.models.deepseek_mtp import (
@@ -36,7 +35,6 @@ from vllm.model_executor.models.deepseek_mtp import (
     SharedHead)
 from vllm.model_executor.models.deepseek_v2 import DeepseekV2DecoderLayer
 from vllm.model_executor.models.utils import maybe_prefix
-from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.sequence import IntermediateTensors
 
 
@@ -168,7 +166,7 @@ class CustomDeepSeekMultiTokenPredictor(DeepSeekMultiTokenPredictor):
     def compute_logits(
         self,
         hidden_states: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
+        sampling_metadata,  # type: ignore
         spec_step_idx: int = 0,
     ) -> torch.Tensor:
         current_step_idx = (spec_step_idx % self.num_mtp_layers)
@@ -188,8 +186,6 @@ class CustomDeepSeekMTP(DeepSeekMTP):
                                                        prefix=maybe_prefix(
                                                            prefix, "model"))
 
-        self.sampler = get_sampler()
-
     def forward(
         self,
         input_ids: torch.Tensor,
@@ -204,4 +200,4 @@ class CustomDeepSeekMTP(DeepSeekMTP):
         hidden_states = self.model(input_ids, positions, kv_caches,
                                    attn_metadata, previous_hidden_states,
                                    inputs_embeds, spec_step_idx)
-        return hidden_states
+        return hidden_states

vllm_ascend/models/deepseek_v2.py

@@ -479,4 +479,8 @@ class CustomDeepseekV2ForCausalLM(DeepseekV2ForCausalLM):
         return loaded_params
 
 
+class CustomDeepseekV3ForCausalLM(CustomDeepseekV2ForCausalLM):
+    pass
+
+
 DeepseekV2DecoderLayer.__init__ = CustomDeepseekV2DecoderLayer.__init__

vllm_ascend/models/deepseek_v3.py

@@ -1,27 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
-# Copyright 2023 The vLLM team.
-# Copyright 2023 DeepSeek-AI and the HuggingFace Inc. team. All rights reserved.
-#
-# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
-# and OPT implementations in this library. It has been modified from its
-# original forms to accommodate minor architectural differences compared
-# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from vllm_ascend.models.deepseek_v2 import CustomDeepseekV2ForCausalLM
-
-
-class CustomDeepseekV3ForCausalLM(CustomDeepseekV2ForCausalLM):
-    pass

vllm_ascend/models/qwen3_next.py

@@ -50,7 +50,6 @@ from vllm.model_executor.models.utils import (
     AutoWeightsLoader, PPMissingLayer, extract_layer_index,
     is_pp_missing_parameter, make_empty_intermediate_tensors_factory,
     make_layers, maybe_prefix)
-from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.model_executor.utils import set_weight_attrs
 from vllm.platforms import current_platform
 from vllm.sequence import IntermediateTensors
@@ -1079,9 +1078,9 @@ class Qwen3NextForCausalLM(nn.Module, HasInnerState, SupportsLoRA, SupportsPP,
             use_v1=True)
 
     def compute_logits(
-        self,
-        hidden_states: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
+            self,
+            hidden_states: torch.Tensor,
+            sampling_metadata,  # type: ignore
     ) -> Optional[torch.Tensor]:
         return self.logits_processor(self.lm_head, hidden_states,
                                      sampling_metadata)