fp16 inference

model.py

@@ -53,9 +53,9 @@ 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":
@@ -88,7 +88,7 @@ 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()