fix device issue in single gpu case

Author: SunMarc

src/diffusers/models/modeling_utils.py

@@ -920,14 +920,12 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
                 else:  # else let accelerate handle loading and dispatching.
                     # Load weights and dispatch according to the device_map
                     # by default the device_map is None and the weights are loaded on the CPU
-                    force_hook = True
                     device_map = _determine_device_map(
                         model, device_map, max_memory, torch_dtype, keep_in_fp32_modules, hf_quantizer
                     )
                     if device_map is None and is_sharded:
                         # we load the parameters on the cpu
                         device_map = {"": "cpu"}
-                        force_hook = False
                     try:
                         accelerate.load_checkpoint_and_dispatch(
                             model,
@@ -937,7 +935,6 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
                             offload_folder=offload_folder,
                             offload_state_dict=offload_state_dict,
                             dtype=torch_dtype,
-                            force_hooks=force_hook,
                             strict=True,
                         )
                     except AttributeError as e:
@@ -967,7 +964,6 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
                                 offload_folder=offload_folder,
                                 offload_state_dict=offload_state_dict,
                                 dtype=torch_dtype,
-                                force_hooks=force_hook,
                                 strict=True,
                             )
                             model._undo_temp_convert_self_to_deprecated_attention_blocks()

src/diffusers/pipelines/pipeline_utils.py

@@ -411,6 +411,13 @@ def module_is_offloaded(module):
         pipeline_is_sequentially_offloaded = any(
             module_is_sequentially_offloaded(module) for _, module in self.components.items()
         )
+
+        is_pipeline_device_mapped = self.hf_device_map is not None and len(self.hf_device_map) > 1
+        if is_pipeline_device_mapped:
+            raise ValueError(
+                "It seems like you have activated a device mapping strategy on the pipeline which doesn't allow explicit device placement using `to()`. You can call `reset_device_map()` first and then call `to()`."
+            )
+
         if device and torch.device(device).type == "cuda":
             if pipeline_is_sequentially_offloaded and not pipeline_has_bnb:
                 raise ValueError(
@@ -422,12 +429,6 @@ def module_is_offloaded(module):
                     "You are trying to call `.to('cuda')` on a pipeline that has models quantized with `bitsandbytes`. Your current `accelerate` installation does not support it. Please upgrade the installation."
                 )
 
-        is_pipeline_device_mapped = self.hf_device_map is not None and len(self.hf_device_map) > 1
-        if is_pipeline_device_mapped:
-            raise ValueError(
-                "It seems like you have activated a device mapping strategy on the pipeline which doesn't allow explicit device placement using `to()`. You can call `reset_device_map()` first and then call `to()`."
-            )
-
         # Display a warning in this case (the operation succeeds but the benefits are lost)
         pipeline_is_offloaded = any(module_is_offloaded(module) for _, module in self.components.items())
         if pipeline_is_offloaded and device and torch.device(device).type == "cuda":