[5/N] pass the whole config to model (#9983)

Signed-off-by: youkaichao <youkaichao@gmail.com>

vllm/model_executor/models/utils.py

@@ -11,11 +11,8 @@ from transformers import PretrainedConfig
 import vllm.envs as envs
 from vllm.attention.selector import (_Backend, backend_name_to_enum,
                                      get_global_forced_attn_backend)
-from vllm.config import (CacheConfig, LoRAConfig, MultiModalConfig,
-                         SchedulerConfig)
+from vllm.config import VllmConfig
 from vllm.logger import init_logger
-from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.model_loader.loader import build_model
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.model_executor.models import ModelRegistry
 from vllm.multimodal.base import MultiModalPlaceholderMap, NestedTensors
@@ -236,12 +233,7 @@ class AutoWeightsLoader:
 
 def init_vllm_registered_model(
     hf_config: PretrainedConfig,
-    cache_config: Optional[CacheConfig],
-    quant_config: Optional[QuantizationConfig],
-    *,
-    lora_config: Optional[LoRAConfig] = None,
-    multimodal_config: Optional[MultiModalConfig] = None,
-    scheduler_config: Optional[SchedulerConfig] = None,
+    vllm_config: VllmConfig,
     prefix: str = "",
 ) -> nn.Module:
     """
@@ -249,18 +241,11 @@ def init_vllm_registered_model(
     based on the arguments passed to the outer vLLM model.
     """
     model_class, _ = ModelRegistry.resolve_model_cls(hf_config.architectures)
+    import copy
+    copied_config = copy.deepcopy(vllm_config)
+    copied_config.model_config.hf_config = hf_config
 
-    return build_model(
-        model_class,
-        None,
-        hf_config,
-        cache_config,
-        quant_config,
-        lora_config=lora_config,
-        multimodal_config=multimodal_config,
-        scheduler_config=scheduler_config,
-        prefix=prefix,
-    )
+    return model_class(vllm_config=copied_config, prefix=prefix)
 
 
 @overload