Load model only once.

app.py

@@ -11,6 +11,8 @@ image_width = 256
 image_height = 192
 samples_dir = "/tmp"
 
+pipe = DiffusionPipeline.from_pretrained(model, safety_checker=None, requires_safety_checker=False)
+
 
 # Borrowed from here: https://stackoverflow.com/a/73667318
 def quantize_to_palette(_image, _palette):
@@ -58,9 +60,9 @@ def palette_to_attr(_palette):
 
 
 def generate(prompt, seed):
-    pipe = DiffusionPipeline.from_pretrained(model, safety_checker=None, requires_safety_checker=False)
     generator = torch.Generator("cpu").manual_seed(int(seed))
-    raw_image = pipe(prompt, height=image_height, width=image_width, num_inference_steps=20, generator=generator).images[0]
+    raw_image = \
+    pipe(prompt, height=image_height, width=image_width, num_inference_steps=20, generator=generator).images[0]
     palette = np.array(
         [[0, 0, 0], [0, 0, 255], [0, 255, 0], [0, 255, 255], [255, 0, 0], [255, 0, 255], [255, 255, 0],
          [255, 255, 255]])