optimize sd/paint_by_example modle VRAM usage

lama_cleaner/model/paint_by_example.py

@@ -22,16 +22,30 @@ class PaintByExample(InpaintModel):
         use_gpu = device == torch.device('cuda') and torch.cuda.is_available()
         torch_dtype = torch.float16 if use_gpu and fp16 else torch.float32
         model_kwargs = {"local_files_only": kwargs.get('local_files_only', False)}
+
+        if kwargs['disable_nsfw'] or kwargs.get('cpu_offload', False):
+            logger.info("Disable Paint By Example Model NSFW checker")
+            model_kwargs.update(dict(
+                safety_checker=None,
+                requires_safety_checker=False
+            ))
+
         self.model = DiffusionPipeline.from_pretrained(
             "Fantasy-Studio/Paint-by-Example",
             torch_dtype=torch_dtype,
             **model_kwargs
         )
-        self.model = self.model.to(device)
+
         self.model.enable_attention_slicing()
+        if kwargs.get('enable_xformers', False):
+            self.model.enable_xformers_memory_efficient_attention()
+
         # TODO: gpu_id
-        if kwargs.get('cpu_offload', False) and torch.cuda.is_available():
+        if kwargs.get('cpu_offload', False) and use_gpu:
+            self.model.image_encoder = self.model.image_encoder.to(device)
             self.model.enable_sequential_cpu_offload(gpu_id=0)
+        else:
+            self.model = self.model.to(device)
 
     def forward(self, image, mask, config: Config):
         """Input image and output image have same size