Update app.py

app.py

@@ -1,36 +1,51 @@
 import gradio as gr
 from diffusers import DiffusionPipeline
-import random
+import torch
+#für die komplexere Variante der Erzeugung
+#from diffusers import DDPMScheduler, UNet2DModel
+#from PIL import Image
+#import numpy as np
 
 
-#########################
+#######################################
 #Alternativ erzeugen
 #gr.Interface.load("models/stabilityai/stable-diffusion-2").launch()
 
-pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2")
 
-def erzeuge():
+#######################################
+#Bild nach dem eingegebenen prompt erzeugen - mit Pipeline
+def erzeuge(prompt):
     pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2")
-    images = [
-        (random.choice(
-            [
-                "https://images.unsplash.com/photo-1507003211169-0a1dd7228f2d?ixlib=rb-1.2.1&ixid=MnwxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8&auto=format&fit=crop&w=387&q=80",
-                "https://images.unsplash.com/photo-1554151228-14d9def656e4?ixlib=rb-1.2.1&ixid=MnwxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8&auto=format&fit=crop&w=386&q=80",
-                "https://images.unsplash.com/photo-1542909168-82c3e7fdca5c?ixlib=rb-1.2.1&ixid=MnwxMjA3fDB8MHxzZWFyY2h8MXx8aHVtYW4lMjBmYWNlfGVufDB8fDB8fA%3D%3D&w=1000&q=80",
-                "https://images.unsplash.com/photo-1546456073-92b9f0a8d413?ixlib=rb-1.2.1&ixid=MnwxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8&auto=format&fit=crop&w=387&q=80",
-                "https://images.unsplash.com/photo-1601412436009-d964bd02edbc?ixlib=rb-1.2.1&ixid=MnwxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8&auto=format&fit=crop&w=464&q=80",
-            ]
-        ), f"label {i}" if i != 0 else "label" * 50)
-        for i in range(3)
-    ]
-    return images
+    pipeline.to("cuda")
+    return pipeline(prompt).images[0]
 
+########################################
+#Bild erzeugen - nich über Pipeline sondern mit mehr Einstellungsmöglichkeiten
+def erzeuge_komplex(prompt):    
+    scheduler = DDPMScheduler.from_pretrained("google/ddpm-cat-256")
+    model = UNet2DModel.from_pretrained("google/ddpm-cat-256").to("cuda")
+    scheduler.set_timesteps(50)
+    
+    sample_size = model.config.sample_size
+    noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda")
+    input = noise
+    
+    for t in scheduler.timesteps:
+        with torch.no_grad():
+            noisy_residual = model(input, t).sample
+            prev_noisy_sample = scheduler.step(noisy_residual, t, input).prev_sample
+            input = prev_noisy_sample
+    
+    image = (input / 2 + 0.5).clamp(0, 1)
+    image = image.cpu().permute(0, 2, 3, 1).numpy()[0]
+    image = Image.fromarray((image * 255).round().astype("uint8"))
+    return image
 
 
 with gr.Blocks() as demo:
     with gr.Column(variant="panel"):
         with gr.Row(variant="compact"):
-            text = gr.Textbox(
+            user_input = gr.Textbox(
                 label="Deine Beschreibung:",
                 show_label=False,
                 max_lines=1,
@@ -44,7 +59,7 @@ with gr.Blocks() as demo:
             label="Erzeugte Bilder", show_label=False, elem_id="gallery"
         ).style(columns=[2], rows=[2], object_fit="contain", height="auto")
 
-    btn.click(erzeuge, None, gallery)
+    btn.click(erzeuge, inputs=[user_input], gallery)
 
 if __name__ == "__main__":
     demo.launch()