Update app.py

app.py

@@ -1,146 +1,74 @@
 import gradio as gr
-import numpy as np
-import random
-from diffusers import DiffusionPipeline
 import torch
+# region_offset = torch.tensor(region_offset).int()
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
+from utils import gen_image_as_per_prompt
 
-if torch.cuda.is_available():
-    torch.cuda.max_memory_allocated(device=device)
-    pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, variant="fp16", use_safetensors=True)
-    pipe.enable_xformers_memory_efficient_attention()
-    pipe = pipe.to(device)
-else: 
-    pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", use_safetensors=True)
-    pipe = pipe.to(device)
+styles = ["depthmap", "cosmicgalaxy", "concept-art", "Marc Allante", "midjourney-style", "No style"]
+styleValues = ["learned_embeds_depthmap.bin",
+               "learned_embeds_cosmic-galaxy-characters-style.bin",
+               "learned_embeds_sd_concept-art.bin",
+               "learned_embeds_style-of-marc-allante.bin",
+               "learned_embeds_midjourney.bin",
+               ""]
+seed_values = [30, 24, 35, 47, 78, 42]
 
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
+styles_dict = dict(zip(styles, styleValues))
+seed_dict = dict(zip(styles, seed_values))
 
-def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
 
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-        
-    generator = torch.Generator().manual_seed(seed)
-    
-    image = pipe(
-        prompt = prompt, 
-        negative_prompt = negative_prompt,
-        guidance_scale = guidance_scale, 
-        num_inference_steps = num_inference_steps, 
-        width = width, 
-        height = height,
-        generator = generator
-    ).images[0] 
-    
-    return image
+# Custom loss function
+def reduce_highlight(images):
+    """Calculates the mean absolute error for amber color.
+
+    Args:
+      images: A tensor of shape (batch_size, channels, height, width).
+      target_red: Target red value for amber.
+      target_green: Target green value for amber.
+      target_blue: Target blue value for amber.
 
-examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
-]
+    Returns:
+      The mean absolute error.
+      #target_red=0.8, target_green=0.6, target_blue=0.4
+    """
 
-css="""
-#col-container {
-    margin: 0 auto;
-    max-width: 520px;
-}
-"""
+    red_error = torch.abs(images[:, 0] - 0.12).mean()
+    green_error = torch.abs(images[:, 1] - 0.2).mean()
+    blue_error = torch.abs(images[:, 2] - 0.15).mean()
 
-if torch.cuda.is_available():
-    power_device = "GPU"
-else:
-    power_device = "CPU"
+    # You can adjust weights for each channel if needed
+    amber_error = (red_error + green_error + blue_error) / 3
+    return amber_error
 
-with gr.Blocks(css=css) as demo:
-    
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown(f"""
-        # Text-to-Image Gradio Template
-        Currently running on {power_device}.
-        """)
-        
-        with gr.Row():
-            
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-            
-            run_button = gr.Button("Run", scale=0)
-        
-        result = gr.Image(label="Result", show_label=False)
+def _inference(text, style, use_loss=False):
+    if use_loss:
+        image = gen_image_as_per_prompt(text, styles_dict[style], seed_dict[style], reduce_highlight)
+    else:
+        image = gen_image_as_per_prompt(text, styles_dict[style], seed_dict[style])
+    return image
 
-        with gr.Accordion("Advanced Settings", open=False):
-            
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
-            )
-            
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-            
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-            
-            with gr.Row():
-                
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=512,
-                )
-                
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=512,
-                )
-            
-            with gr.Row():
-                
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=10.0,
-                    step=0.1,
-                    value=0.0,
-                )
-                
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=12,
-                    step=1,
-                    value=2,
-                )
-        
-        gr.Examples(
-            examples = examples,
-            inputs = [prompt]
-        )
 
-    run_button.click(
-        fn = infer,
-        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
-        outputs = [result]
-    )
+title = "Stable Diffusion with different styles"
+description = "In this demo, the word 'puppy' is replaced by the style that is selected"
+examples = [["oil painting of a dragon in puppy style", "mosiac", True],
+            ["Spiderman in puppy style", "midjourney", True],
+            ["Batman in puppy style", "matrix", False],
+             ["Mojo Jojo in puppy style", "No style", False]]
 
-demo.queue().launch()
+demo = gr.Interface(
+    _inference,
+    inputs=[
+        gr.Textbox(placeholder="Type a prompt with word 'puppy' in it..", container=False, scale=7),
+        gr.Radio(styles, label="Select a Style"),
+        gr.Checkbox(label="Use custom loss")
+    ],
+    outputs=[
+        gr.Image(width=256, height=256, label="output")
+        # gr.Text(label="output")
+    ],
+    title=title,
+    description=description,
+    examples=examples,
+    cache_examples=False
+)
+demo.launch(debug=True)