Update app.py

app.py

@@ -1,146 +1,98 @@
 import gradio as gr
-import numpy as np
-import random
-from diffusers import DiffusionPipeline
 import torch
+from omegaconf import OmegaConf
+from gligen.task_grounded_generation import grounded_generation_box, load_ckpt, load_common_ckpt
+
+import json
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+from functools import partial
+from collections import Counter
+import math
+import gc
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
+from gradio import processing_utils
+from typing import Optional
 
-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)
+import warnings
 
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
+from datetime import datetime
 
-def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
+from huggingface_hub import hf_hub_download
+hf_hub_download = partial(hf_hub_download, library_name="gligen_demo")
 
-    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
+import sys
 
-examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
-]
+class ImageMask(gr.components.Image):
+    is_template = True
+    def __init__(self, **kwargs):
+        super().__init__(source="upload", tool="sketch", interactive=True, **kwargs)
 
-css="""
-#col-container {
-    margin: 0 auto;
-    max-width: 520px;
-}
-"""
+    def preprocess(self, x):
+        if x is None:
+            return x
+        if self.tool == "sketch" and self.source in ["upload", "webcam"] and type(x) != dict:
+            decode_image = processing_utils.decode_base64_to_image(x)
+            width, height = decode_image.size
+            mask = np.zeros((height, width, 4), dtype=np.uint8)
+            mask[..., -1] = 255
+            mask = self.postprocess(mask)
+            x = {'image': x, 'mask': mask}
+        return super().preprocess(x)
 
-if torch.cuda.is_available():
-    power_device = "GPU"
-else:
-    power_device = "CPU"
 
-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)
+with gr.Blocks(
+    analytics_enabled=False,
+    title="GLIGen demo",
+) as main:
+  with gr.Row():
+        with gr.Column(scale=4):
+            sketch_pad_trigger = gr.Number(value=0, visible=False)
+            sketch_pad_resize_trigger = gr.Number(value=0, visible=False)
+            init_white_trigger = gr.Number(value=0, visible=False)
+            image_scale = gr.Number(value=0, elem_id="image_scale", visible=False)
+            new_image_trigger = gr.Number(value=0, visible=False)
 
-        with gr.Accordion("Advanced Settings", open=False):
-            
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
+            task = gr.Radio(
+                choices=["Grounded Generation", 'Grounded Inpainting'],
+                type="value",
+                value="Grounded Generation",
+                label="Task",
             )
-            
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
+            language_instruction = gr.Textbox(
+                label="Language instruction",
             )
-            
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-            
+            grounding_instruction = gr.Textbox(
+                label="Grounding instruction (Separated by semicolon)",
+            )
+            with gr.Row():
+                sketch_pad = ImageMask(label="Sketch Pad", elem_id="img2img_image")
+                out_imagebox = gr.Image(type="pil", label="Parsed Sketch Pad")
+            with gr.Row():
+                clear_btn = gr.Button(value='Clear')
+                gen_btn = gr.Button(value='Generate')
+            with gr.Accordion("Advanced Options", open=False):
+                with gr.Column():
+                    alpha_sample = gr.Slider(minimum=0, maximum=1.0, step=0.1, value=0.3, label="Scheduled Sampling (τ)")
+                    guidance_scale = gr.Slider(minimum=0, maximum=50, step=0.5, value=7.5, label="Guidance Scale")
+                    batch_size = gr.Slider(minimum=1, maximum=4, step=1, value=2, label="Number of Samples")
+                    append_grounding = gr.Checkbox(value=True, label="Append grounding instructions to the caption")
+                    use_actual_mask = gr.Checkbox(value=False, label="Use actual mask for inpainting", visible=False)
+                    with gr.Row():
+                        fix_seed = gr.Checkbox(value=True, label="Fixed seed")
+                        rand_seed = gr.Slider(minimum=0, maximum=1000, step=1, value=0, label="Seed")
+                    with gr.Row(visible=False):
+                        use_style_cond = gr.Checkbox(value=False, label="Enable Style Condition")
+                        style_cond_image = gr.Image(type="pil", label="Style Condition", visible=False, interactive=True)
+        with gr.Column(scale=4):
+            gr.HTML('<span style="font-size: 20px; font-weight: bold">Generated Images</span>')
             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,
-                )
-            
+                out_gen_1 = gr.Image(type="pil", visible=True, show_label=False)
+                out_gen_2 = gr.Image(type="pil", visible=True, show_label=False)
             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]
-        )
+                out_gen_3 = gr.Image(type="pil", visible=False, show_label=False)
+                out_gen_4 = gr.Image(type="pil", visible=False, show_label=False)
 
-    run_button.click(
-        fn = infer,
-        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
-        outputs = [result]
-    )
+        state = gr.State({})
 
-demo.queue().launch()
+main.queue(concurrency_count=1, api_open=False).launch(share=False, show_api=False, show_error=True)