[CI] Upgrade vllm to newest commit (#3182)

### What this PR does / why we need it?
Upgrade vLLM to newest commit

- Fix the aclgraph doesn't work problem, caused by
24fab45d96
- Fix PoolerOutput import error, caused by
755ed7b05b
- Fix the aclgraph weight load error to keep the same with torchair fix.
4492e3a554

### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
All test should pass


- vLLM version: v0.10.2
- vLLM main:
52d0cb8458

---------

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

vllm_ascend/quantization/quant_config.py

@@ -33,7 +33,6 @@ from vllm.model_executor.layers.quantization.base_config import (
 from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     UnquantizedEmbeddingMethod, VocabParallelEmbedding)
-from vllm.model_executor.parameter import PerTensorScaleParameter
 from vllm.model_executor.utils import set_weight_attrs
 
 from vllm_ascend.distributed.parallel_state import (get_mlp_tp_group,
@@ -251,7 +250,6 @@ class AscendLinearMethod(LinearMethodBase):
         **extra_weight_attrs,
     ) -> None:
         output_size_per_partition = sum(output_partition_sizes)
-        weight_loader = extra_weight_attrs.get("weight_loader")
 
         weight_dict = self.quant_method.get_weight(input_size_per_partition,
                                                    output_size_per_partition,
@@ -264,8 +262,7 @@ class AscendLinearMethod(LinearMethodBase):
 
         pertensor_dict = self.quant_method.get_pertensor_param(params_dtype)
         for pertensor_name, pertensor_param in pertensor_dict.items():
-            param = PerTensorScaleParameter(data=pertensor_param,
-                                            weight_loader=weight_loader)
+            param = torch.nn.Parameter(pertensor_param, requires_grad=False)
             # disable warning
             param.ignore_warning = True
             layer.register_parameter(pertensor_name, param)