Upload 3 files

app.py

@@ -6,107 +6,34 @@ import warnings
 warnings.filterwarnings('ignore')
 
 # Constants
-MODEL_ID = "CompVis/stable-diffusion-v1-4"  # Using v1.4 which is publicly available
-DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
-DTYPE = torch.float32  # Using float32 for better compatibility
+MODEL_ID = "google/ddpm-celebahq-256"  # Using a simpler model
+DEVICE = "cpu"  # Force CPU for better compatibility
+DTYPE = torch.float32
 
-class FooocusGenerator:
-    def __init__(self):
-        self.pipe = None
-        self.load_models()
-
-    def load_models(self):
-        try:
-            # Initialize pipeline with basic settings
-            self.pipe = StableDiffusionPipeline.from_pretrained(
-                MODEL_ID,
-                torch_dtype=DTYPE,
-                use_auth_token=False  # No auth token needed for v1.4
-            )
-            
-            if DEVICE == "cuda":
-                self.pipe.enable_attention_slicing()
-            self.pipe = self.pipe.to(DEVICE)
-            
-        except Exception as e:
-            print(f"Error loading model: {str(e)}")
-            raise e
-
-    def generate_image(
-        self,
-        prompt,
-        negative_prompt,
-        style_selections,
-        performance_selection,
-        aspect_ratios_selection,
-        image_number,
-        image_seed,
-        sharpness,
-        guidance_scale,
-        progress=gr.Progress()
-    ):
-        try:
-            # Process style selections
-            processed_prompt = self.process_style(prompt, style_selections)
-            
-            # Set seed for reproducibility
-            if image_seed == -1:
-                image_seed = torch.randint(0, 2147483647, (1,)).item()
-            
-            generator = torch.manual_seed(image_seed)
-            
-            # Set steps based on performance selection
-            steps = {
-                "Speed": 20,
-                "Quality": 30,
-                "Extreme Speed": 15
-            }.get(performance_selection, 30)
-            
-            # Set image dimensions based on aspect ratio
-            dimensions = {
-                "Square": (512, 512),
-                "Portrait": (512, 768),
-                "Landscape": (768, 512)
-            }.get(aspect_ratios_selection, (512, 512))
-            
-            # Generate image with basic settings
-            with torch.inference_mode():
-                image = self.pipe(
-                    prompt=processed_prompt,
-                    negative_prompt=negative_prompt,
-                    num_inference_steps=steps,
-                    guidance_scale=guidance_scale,
-                    height=dimensions[1],
-                    width=dimensions[0],
-                    generator=generator
-                ).images[0]
-            
-            return image
-            
-        except Exception as e:
-            print(f"Error generating image: {str(e)}")
-            return None
-
-    def process_style(self, prompt, style_selections):
-        style_modifiers = {
-            "Fooocus V2": ", professional, high quality, detailed",
-            "Fooocus Enhance": ", enhanced details, perfect composition",
-            "Fooocus Sharp": ", sharp focus, high resolution",
-            "Fooocus Masterpiece": ", masterpiece, best quality, award winning",
-        }
+def generate_image(prompt, negative_prompt, guidance_scale=7.5, steps=30):
+    try:
+        # Initialize pipeline with basic settings
+        pipe = StableDiffusionPipeline.from_pretrained(MODEL_ID)
+        pipe = pipe.to(DEVICE)
         
-        selected_modifiers = [style_modifiers[style] for style in style_selections if style in style_modifiers]
-        processed_prompt = prompt + " ".join(selected_modifiers)
+        # Generate image
+        with torch.inference_mode():
+            image = pipe(
+                prompt=prompt,
+                negative_prompt=negative_prompt,
+                num_inference_steps=steps,
+                guidance_scale=guidance_scale,
+            ).images[0]
         
-        return processed_prompt
-
-# Initialize the generator
-generator = FooocusGenerator()
+        return image
+    except Exception as e:
+        print(f"Error generating image: {str(e)}")
+        return None
 
 # Create the Gradio interface
-with gr.Blocks(title="Fooocus Web") as demo:
-    gr.Markdown("# 🎨 Fooocus Web")
-    gr.Markdown("Generate high-quality images with advanced controls")
+with gr.Blocks(title="Simple Image Generator") as demo:
+    gr.Markdown("# 🎨 Simple Image Generator")
+    gr.Markdown("Generate images with text prompts")
     
     with gr.Row():
         with gr.Column():
@@ -119,53 +46,16 @@ with gr.Blocks(title="Fooocus Web") as demo:
                 label="Negative Prompt",
                 placeholder="What you don't want in the image...",
                 info="Specify unwanted elements",
-                value="ugly, blurry, low quality, distorted, deformed"
+                value="ugly, blurry, low quality"
             )
             
             with gr.Row():
-                style_selections = gr.CheckboxGroup(
-                    choices=["Fooocus V2", "Fooocus Enhance", "Fooocus Sharp", "Fooocus Masterpiece"],
-                    label="Style Selections",
-                    value=["Fooocus V2"]
-                )
-            
-            with gr.Row():
-                performance_selection = gr.Radio(
-                    choices=["Speed", "Quality", "Extreme Speed"],
-                    label="Performance",
-                    value="Quality"
-                )
-                
-                aspect_ratios_selection = gr.Radio(
-                    choices=["Square", "Portrait", "Landscape"],
-                    label="Aspect Ratio",
-                    value="Square"
-                )
-            
-            with gr.Row():
-                image_number = gr.Slider(
-                    minimum=1,
-                    maximum=32,
-                    value=1,
+                steps = gr.Slider(
+                    minimum=10,
+                    maximum=50,
+                    value=30,
                     step=1,
-                    label="Number of Images"
-                )
-                
-                image_seed = gr.Slider(
-                    minimum=-1,
-                    maximum=2147483647,
-                    step=1,
-                    value=-1,
-                    label="Seed (-1 for random)"
-                )
-            
-            with gr.Row():
-                sharpness = gr.Slider(
-                    minimum=0.0,
-                    maximum=30.0,
-                    value=2.0,
-                    step=0.1,
-                    label="Sharpness"
+                    label="Steps"
                 )
                 
                 guidance_scale = gr.Slider(
@@ -182,17 +72,12 @@ with gr.Blocks(title="Fooocus Web") as demo:
             output_image = gr.Image(label="Generated Image", type="pil")
             
     generate_btn.click(
-        fn=generator.generate_image,
+        fn=generate_image,
         inputs=[
             prompt,
             negative_prompt,
-            style_selections,
-            performance_selection,
-            aspect_ratios_selection,
-            image_number,
-            image_seed,
-            sharpness,
             guidance_scale,
+            steps,
         ],
         outputs=output_image
     )

requirements.txt

@@ -1,12 +1,8 @@
 torch==1.13.1
 torchvision==0.14.1
-diffusers==0.3.0
-transformers==4.19.2
+diffusers==0.2.4
+transformers==4.18.0
 accelerate==0.12.0
-safetensors==0.3.1
 gradio==3.16.2
 opencv-python==4.8.0.74
 einops==0.6.1
-pytorch-lightning==1.9.0
-omegaconf==2.3.0
-huggingface-hub>=0.19