feat(config): support parsing torch.dtype (#1641)

Signed-off-by: Aaron <29749331+aarnphm@users.noreply.github.com>

vllm/config.py

-from typing import Optional
+from typing import Optional, Union
 
 import torch
 from transformers import PretrainedConfig
@@ -58,7 +58,7 @@ class ModelConfig:
         trust_remote_code: bool,
         download_dir: Optional[str],
         load_format: str,
-        dtype: str,
+        dtype: Union[str, torch.dtype],
         seed: int,
         revision: Optional[str] = None,
         tokenizer_revision: Optional[str] = None,
@@ -331,7 +331,7 @@ _STR_DTYPE_TO_TORCH_DTYPE = {
 
 def _get_and_verify_dtype(
     config: PretrainedConfig,
-    dtype: str,
+    dtype: Union[str, torch.dtype],
 ) -> torch.dtype:
     # NOTE: getattr(config, "torch_dtype", torch.float32) is not correct
     # because config.torch_dtype can be None.
@@ -339,10 +339,12 @@ def _get_and_verify_dtype(
     if config_dtype is None:
         config_dtype = torch.float32
 
+    if isinstance(dtype, str):
         dtype = dtype.lower()
         if dtype == "auto":
             if config_dtype == torch.float32:
-            # Following the common practice, we use float16 for float32 models.
+                # Following the common practice, we use float16 for float32
+                # models.
                 torch_dtype = torch.float16
             else:
                 torch_dtype = config_dtype
@@ -350,6 +352,10 @@ def _get_and_verify_dtype(
             if dtype not in _STR_DTYPE_TO_TORCH_DTYPE:
                 raise ValueError(f"Unknown dtype: {dtype}")
             torch_dtype = _STR_DTYPE_TO_TORCH_DTYPE[dtype]
+    elif isinstance(dtype, torch.dtype):
+        torch_dtype = dtype
+    else:
+        raise ValueError(f"Unknown dtype: {dtype}")
 
     # Verify the dtype.
     if torch_dtype != config_dtype: