[CPU offload] correct cpu offload (#1968)

* [CPU offload] correct cpu offload

* [CPU offload] correct cpu offload

* finish

* finish

* Update docs/source/en/optimization/fp16.mdx
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

* Update src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

docs/source/en/optimization/fp16.mdx

@@ -149,7 +149,7 @@ You may see a small performance boost in VAE decode on multi-image batches. Ther
 
 ## Offloading to CPU with accelerate for memory savings
 
-For additional memory savings, you can offload the weights to CPU and load them to GPU when performing the forward pass.
+For additional memory savings, you can offload the weights to CPU and only load them to GPU when performing the forward pass.
 
 To perform CPU offloading, all you have to do is invoke [`~StableDiffusionPipeline.enable_sequential_cpu_offload`]:
 
@@ -162,16 +162,15 @@ pipe = StableDiffusionPipeline.from_pretrained(
     
     torch_dtype=torch.float16,
 )
-pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
 pipe.enable_sequential_cpu_offload()
 image = pipe(prompt).images[0]
 ```
 
-And you can get the memory consumption to < 2GB.
+And you can get the memory consumption to < 3GB.
 
-If is also possible to chain it with attention slicing for minimal memory consumption, running it in as little as < 800mb of GPU vRAM:
+If is also possible to chain it with attention slicing for minimal memory consumption (< 2GB).
 
 ```Python
 import torch
@@ -182,7 +181,6 @@ pipe = StableDiffusionPipeline.from_pretrained(
     
     torch_dtype=torch.float16,
 )
-pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
 pipe.enable_sequential_cpu_offload()
@@ -191,6 +189,8 @@ pipe.enable_attention_slicing(1)
 image = pipe(prompt).images[0]
 ```
 
+**Note**: When using `enable_sequential_cpu_offload()`, it is important to **not** move the pipeline to CUDA beforehand or else the gain in memory consumption will only be minimal. See [this issue](https://github.com/huggingface/diffusers/issues/1934) for more information.
+
 ## Using Channels Last memory format
 
 Channels last memory format is an alternative way of ordering NCHW tensors in memory preserving dimensions ordering. Channels last tensors ordered in such a way that channels become the densest dimension (aka storing images pixel-per-pixel). Since not all operators currently support channels last format it may result in a worst performance, so it's better to try it and see if it works for your model.

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py

@@ -211,13 +211,10 @@ class AltDiffusionPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, device, offload_buffers=True)
 
     @property
     def _execution_device(self):

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -233,13 +233,10 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     def _execution_device(self):

src/diffusers/pipelines/stable_diffusion/pipeline_cycle_diffusion.py

@@ -236,13 +236,10 @@ class CycleDiffusionPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -208,13 +208,10 @@ class StableDiffusionPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     def _execution_device(self):

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -238,13 +238,10 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -272,13 +272,10 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint_legacy.py

@@ -205,13 +205,10 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_k_diffusion.py

@@ -137,13 +137,10 @@ class StableDiffusionKDiffusionPipeline(DiffusionPipeline):
         device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
             cpu_offload(cpu_offloaded_model, device)
 
         if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
 
     @property
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device