set fp32 to avoid errors

model.py

@@ -53,14 +53,16 @@ class Model:
         ):
             return self.pipe
         model_id = CONTROLNET_MODEL_IDS[task_name]
-        controlnet = ControlNetModel.from_pretrained(model_id, torch_dtype=torch.float32)
+        controlnet = ControlNetModel.from_pretrained(model_id, torch_dtype=torch.float16)
         pipe = StableDiffusionControlNetPipeline.from_pretrained(
-            base_model_id, safety_checker=None, controlnet=controlnet, torch_dtype=torch.float32
+            base_model_id, safety_checker=None, controlnet=controlnet, torch_dtype=torch.float16
         )
         pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
         if self.device.type == "cuda":
             pipe.disable_xformers_memory_efficient_attention()
         pipe.to(self.device)
+        pipe.enable_model_cpu_offload()
+        pipe.enable_vae_slicing()
         torch.cuda.empty_cache()
         gc.collect()
         self.base_model_id = base_model_id
@@ -87,7 +89,7 @@ class Model:
         torch.cuda.empty_cache()
         gc.collect()
         model_id = CONTROLNET_MODEL_IDS[task_name]
-        controlnet = ControlNetModel.from_pretrained(model_id, torch_dtype=torch.float32)
+        controlnet = ControlNetModel.from_pretrained(model_id, torch_dtype=torch.float16)
         controlnet.to(self.device)
         torch.cuda.empty_cache()
         gc.collect()