[Model] Add MiMo-V2-Flash support (#30836)

Signed-off-by: Abatom <abzhonghua@gmail.com>
Signed-off-by: Jumiar <liuanqim10@126.com>
Signed-off-by: Zyann7 <zyann7@outlook.com>
Co-authored-by: Jumiar <liuanqim10@126.com>
Co-authored-by: Zyann7 <zyann7@outlook.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>

vllm/config/__init__.py

@@ -18,6 +18,7 @@ from vllm.config.lora import LoRAConfig
 from vllm.config.model import (
     ModelConfig,
     iter_architecture_defaults,
+    str_dtype_to_torch_dtype,
     try_match_architecture_defaults,
 )
 from vllm.config.multimodal import MultiModalConfig
@@ -72,6 +73,7 @@ __all__ = [
     # From vllm.config.model
     "ModelConfig",
     "iter_architecture_defaults",
+    "str_dtype_to_torch_dtype",
     "try_match_architecture_defaults",
     # From vllm.config.multimodal
     "MultiModalConfig",

vllm/config/model.py

@@ -1849,6 +1849,11 @@ _STR_DTYPE_TO_TORCH_DTYPE = {
     "bfloat16": torch.bfloat16,
 }
 
+
+def str_dtype_to_torch_dtype(type: str):
+    return _STR_DTYPE_TO_TORCH_DTYPE.get(type)
+
+
 # model_type -> reason
 _FLOAT16_NOT_SUPPORTED_MODELS = {
     "gemma2": "Numerical instability. Please use bfloat16 or float32 instead.",