Update app.py

app.py

@@ -3,7 +3,7 @@ import gradio as gr
 from PIL import Image
 import cv2
 import qrcode
-import os, random
+import os, random, gc
 import numpy as np
 from transformers import pipeline
 import PIL.Image
@@ -82,10 +82,11 @@ models =[
 
 controlnet = accelerator.prepare(ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-canny", torch_dtype=torch.float32))
 def plex(qr_code_value, text, neg_text, modil, one, two, three):
-    generator = torch.Generator(device="cpu").manual_seed(random.randint(1, 4836923))
+    gc.collect()
     apol=[]
     modal=""+modil+""
     pipe = accelerator.prepare(StableDiffusionControlNetPipeline.from_pretrained(modal, controlnet=controlnet, torch_dtype=torch.float32, use_safetensors=False, safety_checker=None))
+    pipe.unet.to(memory_format=torch.channels_last)
     pipe.scheduler = accelerator.prepare(DPMSolverMultistepScheduler.from_config(pipe.scheduler.config))
     pipe = pipe.to("cpu")
     negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
@@ -103,6 +104,7 @@ def plex(qr_code_value, text, neg_text, modil, one, two, three):
     cannyimage = np.concatenate([cannyimage, cannyimage, cannyimage], axis=2)
     cannyimage = Image.fromarray(cannyimage)
     images = [cannyimage]
+    generator = torch.Generator(device="cpu").manual_seed(random.randint(1, 4836923))
 
     image = pipe(
         [prompt]*2,
@@ -113,7 +115,7 @@ def plex(qr_code_value, text, neg_text, modil, one, two, three):
         negative_prompt=[neg_prompt]*2,
         controlnet_conditioning_scale=three,
     )
-    for i, imge in enumerate(imoge["images"]):
+    for i, imge in enumerate(image["images"]):
         apol.append(imge)
     apol.append(original)
     apol.append(cannyimage)