Remove `torch_dtype` in to() to end deprecation (#6886)

* remove torch_dtype from to()

* remove torch_dtype from usage scripts.

* remove old lora backend

* Revert "remove old lora backend"

This reverts commit adcddf6ba421f847e7da2a0ce57b9456cae43356.

scripts/convert_gligen_to_diffusers.py

@@ -576,6 +576,6 @@ if __name__ == "__main__":
     )
 
     if args.half:
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
 
     pipe.save_pretrained(args.dump_path)

scripts/convert_original_stable_diffusion_to_diffusers.py

@@ -179,7 +179,7 @@ if __name__ == "__main__":
     )
 
     if args.half:
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
 
     if args.controlnet:
         # only save the controlnet model

scripts/convert_zero123_to_diffusers.py

@@ -801,6 +801,6 @@ if __name__ == "__main__":
     )
 
     if args.half:
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
 
     pipe.save_pretrained(args.dump_path, safe_serialization=args.to_safetensors)

src/diffusers/pipelines/pipeline_utils.py

@@ -775,32 +775,10 @@ class DiffusionPipeline(ConfigMixin, PushToHubMixin):
         Returns:
             [`DiffusionPipeline`]: The pipeline converted to specified `dtype` and/or `dtype`.
         """
-
-        torch_dtype = kwargs.pop("torch_dtype", None)
-        if torch_dtype is not None:
-            deprecate("torch_dtype", "0.27.0", "")
-        torch_device = kwargs.pop("torch_device", None)
-        if torch_device is not None:
-            deprecate("torch_device", "0.27.0", "")
-
-        dtype_kwarg = kwargs.pop("dtype", None)
-        device_kwarg = kwargs.pop("device", None)
+        dtype = kwargs.pop("dtype", None)
+        device = kwargs.pop("device", None)
         silence_dtype_warnings = kwargs.pop("silence_dtype_warnings", False)
 
-        if torch_dtype is not None and dtype_kwarg is not None:
-            raise ValueError(
-                "You have passed both `torch_dtype` and `dtype` as a keyword argument. Please make sure to only pass `dtype`."
-            )
-
-        dtype = torch_dtype or dtype_kwarg
-
-        if torch_device is not None and device_kwarg is not None:
-            raise ValueError(
-                "You have passed both `torch_device` and `device` as a keyword argument. Please make sure to only pass `device`."
-            )
-
-        device = torch_device or device_kwarg
-
         dtype_arg = None
         device_arg = None
         if len(args) == 1:
@@ -873,12 +851,12 @@ class DiffusionPipeline(ConfigMixin, PushToHubMixin):
 
             if is_loaded_in_8bit and dtype is not None:
                 logger.warning(
-                    f"The module '{module.__class__.__name__}' has been loaded in 8bit and conversion to {torch_dtype} is not yet supported. Module is still in 8bit precision."
+                    f"The module '{module.__class__.__name__}' has been loaded in 8bit and conversion to {dtype} is not yet supported. Module is still in 8bit precision."
                 )
 
             if is_loaded_in_8bit and device is not None:
                 logger.warning(
-                    f"The module '{module.__class__.__name__}' has been loaded in 8bit and moving it to {torch_dtype} via `.to()` is not yet supported. Module is still on {module.device}."
+                    f"The module '{module.__class__.__name__}' has been loaded in 8bit and moving it to {dtype} via `.to()` is not yet supported. Module is still on {module.device}."
                 )
             else:
                 module.to(device, dtype)

tests/pipelines/animatediff/test_animatediff.py

@@ -218,7 +218,7 @@ class AnimateDiffPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes))
 
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes))
 

tests/pipelines/animatediff/test_animatediff_video2video.py

@@ -224,7 +224,7 @@ class AnimateDiffVideoToVideoPipelineFastTests(PipelineTesterMixin, unittest.Tes
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes))
 
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes))
 

tests/pipelines/audioldm2/test_audioldm2.py

@@ -483,7 +483,7 @@ class AudioLDM2PipelineFastTests(PipelineTesterMixin, unittest.TestCase):
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes.values()))
 
         # Once we send to fp16, all params are in half-precision, including the logit scale
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = {key: component.dtype for key, component in components.items() if hasattr(component, "dtype")}
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes.values()))
 

tests/pipelines/musicldm/test_musicldm.py

@@ -400,7 +400,7 @@ class MusicLDMPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes.values()))
 
         # Once we send to fp16, all params are in half-precision, including the logit scale
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = {key: component.dtype for key, component in components.items() if hasattr(component, "dtype")}
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes.values()))
 

tests/pipelines/pia/test_pia.py

@@ -231,7 +231,7 @@ class PIAPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes))
 
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes))
 

tests/pipelines/stable_video_diffusion/test_stable_video_diffusion.py

@@ -396,7 +396,7 @@ class StableVideoDiffusionPipelineFastTests(PipelineTesterMixin, unittest.TestCa
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes))
 
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = [component.dtype for component in pipe.components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes))
 

tests/pipelines/test_pipelines.py

@@ -1623,7 +1623,7 @@ class PipelineFastTests(unittest.TestCase):
         sd1 = sd.to(torch.float16)
         sd2 = sd.to(None, torch.float16)
         sd3 = sd.to(dtype=torch.float16)
-        sd4 = sd.to(torch_dtype=torch.float16)
+        sd4 = sd.to(dtype=torch.float16)
         sd5 = sd.to(None, dtype=torch.float16)
         sd6 = sd.to(None, torch_dtype=torch.float16)
 

tests/pipelines/test_pipelines_common.py

@@ -716,7 +716,7 @@ class PipelineTesterMixin:
         model_dtypes = [component.dtype for component in components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float32 for dtype in model_dtypes))
 
-        pipe.to(torch_dtype=torch.float16)
+        pipe.to(dtype=torch.float16)
         model_dtypes = [component.dtype for component in components.values() if hasattr(component, "dtype")]
         self.assertTrue(all(dtype == torch.float16 for dtype in model_dtypes))