Update app.py

app.py

@@ -18,7 +18,7 @@ checkpoints = {
     "4-Step" : ["sdxl_lightning_4step_unet.safetensors", 4],
     "8-Step" : ["sdxl_lightning_8step_unet.safetensors", 8],
 }
-
+loaded = None
 
 # Ensure model and scheduler are initialized in GPU-enabled function
 if torch.cuda.is_available():
@@ -49,19 +49,17 @@ if SAFETY_CHECKER:
 # Function 
 @spaces.GPU(enable_queue=True)
 def generate_image(prompt, ckpt):
+    global loaded
     print(prompt, ckpt)
 
     checkpoint = checkpoints[ckpt][0]
-    num_inference_steps = checkpoints[ckpt][1] 
-
-    if num_inference_steps==1:
-        # Ensure sampler uses "trailing" timesteps and "sample" prediction type for 1-step inference.
-        pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", prediction_type="sample")
-    else:
-        # Ensure sampler uses "trailing" timesteps.
-        pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
+    num_inference_steps = checkpoints[ckpt][1]
+
+    if loaded != num_inference_steps:
+        pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", prediction_type="sample" if num_inference_steps==1 else "epsilon")
+        pipe.unet.load_state_dict(load_file(hf_hub_download(repo, checkpoint), device="cuda"))
+        loaded = num_inference_steps
         
-    pipe.unet.load_state_dict(load_file(hf_hub_download(repo, checkpoint), device="cuda"))
     results = pipe(prompt, num_inference_steps=num_inference_steps, guidance_scale=0)
 
     if SAFETY_CHECKER: