cpu offloading: mutli GPU support (#1143)

mutli GPU support

Author: dblunk88

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -178,7 +178,7 @@ def disable_attention_slicing(self):
         # set slice_size = `None` to disable `attention slicing`
         self.enable_attention_slicing(None)
 
-    def enable_sequential_cpu_offload(self):
+    def enable_sequential_cpu_offload(self, gpu_id=0):
         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
@@ -189,7 +189,7 @@ def enable_sequential_cpu_offload(self):
         else:
             raise ImportError("Please install accelerate via `pip install accelerate`")
 
-        device = torch.device("cuda")
+        device = torch.device(f"cuda:{gpu_id}")
 
         for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae, self.safety_checker]:
             if cpu_offloaded_model is not None: