Update space

app.py

@@ -1,61 +1,26 @@
-# import os
-# import sys
-
-# import torch
-# import torchvision.utils as vutils
-
-# from models.clip_encoder import CLIPModel
-# from models.denoising_network import DenoisingNetwork
-# from utils.sde_functions import SDE, add_noise  # Import SDE class
-
-# sys.path.append(os.path.abspath(os.path.join(os.path.dirname(_file_), '..')))
-
-
-# def sample(text, checkpoint_path='checkpoints/model_epoch_50.pth', output_path='generated_images/sample.png', num_steps=1000, guidance_scale=3.5):
-#     device = 'cuda' if torch.cuda.is_available() else 'cpu'
-#     model = DenoisingNetwork().to(device)
-#     model.load_state_dict(torch.load(checkpoint_path))
-#     model.eval()
-
-#     clip_model = CLIPModel()
-#     text_embedding = clip_model.encode_text(text).to(device)
-
-#     # Initialize SDE instance
-#     sde = SDE(noise_type="additive", noise_scale=0.1, guidance_scale=guidance_scale)
-
-#     with torch.no_grad():
-#         # Start with random noise
-#         x = torch.randn(1, 3, 32, 32).to(device)
-#         for t in reversed(range(1, num_steps + 1)):
-#             time_factor = t / num_steps
-#             x = add_noise(x, time_factor)  # Adding noise following the new schedule
-#             x = sde.reverse_sde(x, time_factor, model)  # Call reverse_sde as a method of the SDE instance
-
-#     vutils.save_image(x, output_path, normalize=True)
-#     print(f"Generated image saved at {output_path}")
-
-# if _name_ == "_main_":
-#     sample("A cat riding a skateboard")
-# scripts/sample.py
 import torch
 from diffusers import StableDiffusionPipeline
+import gradio as gr
 
-
-def generate_images(text_prompts, output_dir='generated_images/'):
+def generate_image(prompt):
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model_id = "CompVis/stable-diffusion-v1-4"
     pipe = StableDiffusionPipeline.from_pretrained(model_id).to(device)
-
-    for i, prompt in enumerate(text_prompts):
-        with torch.no_grad():
-            image = pipe(prompt).images[0]
-            image.save(f"{output_dir}/sample_{i}.png")
-            print(f"Generated image {i+1} saved at {output_dir}/sample_{i}.png")
-
+    
+    with torch.no_grad():
+        image = pipe(prompt).images[0]
+    
+    return image
+
+# Define the Gradio interface
+interface = gr.Interface(
+    fn=generate_image, 
+    inputs="text", 
+    outputs="image",
+    title="Stable Diffusion Image Generator",
+    description="Generate images from text prompts using Stable Diffusion."
+)
+
+# Launch the app
 if __name__ == "__main__":
-    prompts = [
-        "A cat riding a skateboard",
-        "A futuristic city skyline at sunset",
-        "A serene mountain landscape in spring"
-    ]
-    generate_images(prompts)
+    interface.launch()

requirements.txt

@@ -1,6 +1,5 @@
-accelerate
-diffusers
-invisible_watermark
-torch
-transformers
-xformers
+torch==2.0.1
+diffusers==0.21.1
+transformers==4.32.1
+gradio==3.32.0
+accelerate==0.22.0