Improve torch_dtype checks (#40808)

* Improve torch_dtype checks

Signed-off-by: Yuanyuan Chen <[email protected]>

* Apply suggestions from code review

---------

Signed-off-by: Yuanyuan Chen <[email protected]>
Co-authored-by: Joao Gante <[email protected]>

Author: cyyever

src/transformers/commands/chat.py

@@ -289,8 +289,14 @@ class ChatArguments:
     def __post_init__(self):
         """Only used for BC `torch_dtype` argument."""
         # In this case only the BC torch_dtype was given
-        if self.torch_dtype is not None and self.dtype == "auto":
-            self.dtype = self.torch_dtype
+        if self.torch_dtype is not None:
+            if self.dtype is None:
+                self.dtype = self.torch_dtype
+            elif self.torch_dtype != self.dtype:
+                raise ValueError(
+                    f"`torch_dtype` {self.torch_dtype} and `dtype` {self.dtype} have different values. `torch_dtype` is deprecated and "
+                    "will be removed in 4.59.0, please set `dtype` instead."
+                )
 
 
 def chat_command_factory(args: Namespace):

src/transformers/commands/serving.py

@@ -457,8 +457,14 @@ class ServeArguments:
     def __post_init__(self):
         """Only used for BC `torch_dtype` argument."""
         # In this case only the BC torch_dtype was given
-        if self.torch_dtype is not None and self.dtype == "auto":
-            self.dtype = self.torch_dtype
+        if self.torch_dtype is not None:
+            if self.dtype is None:
+                self.dtype = self.torch_dtype
+            elif self.torch_dtype != self.dtype:
+                raise ValueError(
+                    f"`torch_dtype` {self.torch_dtype} and `dtype` {self.dtype} have different values. `torch_dtype` is deprecated and "
+                    "will be removed in 4.59.0, please set `dtype` instead."
+                )
 
 
 class ServeCommand(BaseTransformersCLICommand):

src/transformers/pipelines/keypoint_matching.py

@@ -147,7 +147,7 @@ def __call__(
     def preprocess(self, images, timeout=None):
         images = [load_image(image, timeout=timeout) for image in images]
         model_inputs = self.image_processor(images=images, return_tensors=self.framework)
-        model_inputs = model_inputs.to(self.torch_dtype)
+        model_inputs = model_inputs.to(self.dtype)
         target_sizes = [image.size for image in images]
         preprocess_outputs = {"model_inputs": model_inputs, "target_sizes": target_sizes}
         return preprocess_outputs