Add device arg to offloading with combined pipelines (huggingface#7471)

Co-authored-by: Sayak Paul <[email protected]>

Author: Disty0

src/diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_combined.py

@@ -178,16 +178,16 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
         `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -405,26 +405,26 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_model_cpu_offload(self, gpu_id=0):
+    def enable_model_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
         to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
         method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
         `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
         """
-        self.prior_pipe.enable_model_cpu_offload()
-        self.decoder_pipe.enable_model_cpu_offload()
+        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
         `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -653,16 +653,16 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
         `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)

src/diffusers/pipelines/stable_cascade/pipeline_stable_cascade_combined.py

@@ -117,25 +117,25 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_model_cpu_offload(self, gpu_id=0):
+    def enable_model_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
         to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
         method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
         `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
         """
-        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models (`unet`, `text_encoder`, `vae`, and `safety checker` state dicts) to CPU using 🤗
         Accelerate, significantly reducing memory usage. Models are moved to a `torch.device('meta')` and loaded on a
         GPU only when their specific submodule's `forward` method is called. Offloading happens on a submodule basis.
         Memory savings are higher than using `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)

src/diffusers/pipelines/wuerstchen/pipeline_wuerstchen_combined.py

@@ -112,25 +112,25 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_model_cpu_offload(self, gpu_id=0):
+    def enable_model_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
         to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
         method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
         `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
         """
-        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_model_cpu_offload(gpu_id=gpu_id, device=device)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models (`unet`, `text_encoder`, `vae`, and `safety checker` state dicts) to CPU using 🤗
         Accelerate, significantly reducing memory usage. Models are moved to a `torch.device('meta')` and loaded on a
         GPU only when their specific submodule's `forward` method is called. Offloading happens on a submodule basis.
         Memory savings are higher than using `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)