Update app.py

app.py

@@ -1,97 +1,12 @@
 import os
-import random
-import uuid
-import json
+from gradio_client import Client
 
-import gradio as gr
-import numpy as np
-from PIL import Image
-import spaces
-import torch
-from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
-
-# Use environment variables for flexibility
-MODEL_ID = os.getenv("MODEL_ID", "sd-community/sdxl-flash")
+# Используем переменные окружения для гибкости
+MODEL_ID = os.getenv("MODEL_ID", "KingNish/SDXL-Flash")
 MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096"))
-USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE", "0") == "1"
-ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD", "0") == "1"
-BATCH_SIZE = int(os.getenv("BATCH_SIZE", "1"))  # Allow generating multiple images at once
-
-# Determine device and load model outside of function for efficiency
-device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-pipe = StableDiffusionXLPipeline.from_pretrained(
-    MODEL_ID,
-    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-    use_safetensors=True,
-    add_watermarker=False,
-).to(device)
-pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
-
-# Torch compile for potential speedup (experimental)
-if USE_TORCH_COMPILE:
-    pipe.compile()
-
-# CPU offloading for larger RAM capacity (experimental)
-if ENABLE_CPU_OFFLOAD:
-    pipe.enable_model_cpu_offload()
-
-MAX_SEED = np.iinfo(np.int32).max
-
-def save_image(img):
-    unique_name = str(uuid.uuid4()) + ".png"
-    img.save(unique_name)
-    return unique_name
-
-def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-    return seed
+BATCH_SIZE = int(os.getenv("BATCH_SIZE", "1"))  # Позволяет генерировать несколько изображений за один раз
 
-@spaces.GPU(duration=30, queue=False)
-def generate(
-    prompt: str,
-    negative_prompt: str = "",
-    use_negative_prompt: bool = False,
-    seed: int = 1,
-    width: int = 1024,
-    height: int = 1024,
-    guidance_scale: float = 3,
-    num_inference_steps: int = 30,
-    randomize_seed: bool = False,
-    use_resolution_binning: bool = True, 
-    num_images: int = 1,  # Number of images to generate
-    progress=gr.Progress(track_tqdm=True),
-):
-    seed = int(randomize_seed_fn(seed, randomize_seed))
-    generator = torch.Generator(device=device).manual_seed(seed)
-
-    # Improved options handling
-    options = {
-        "prompt": [prompt] * num_images,
-        "negative_prompt": [negative_prompt] * num_images if use_negative_prompt else None,
-        "width": width,
-        "height": height,
-        "guidance_scale": guidance_scale,
-        "num_inference_steps": num_inference_steps,
-        "generator": generator,
-        "output_type": "pil",
-    }
-
-    # Use resolution binning for faster generation with less VRAM usage
-    if use_resolution_binning:
-        options["use_resolution_binning"] = True
-
-    # Generate images potentially in batches
-    images = []
-    for i in range(0, num_images, BATCH_SIZE):
-        batch_options = options.copy()
-        batch_options["prompt"] = options["prompt"][i:i+BATCH_SIZE]
-        if "negative_prompt" in batch_options:
-            batch_options["negative_prompt"] = options["negative_prompt"][i:i+BATCH_SIZE]
-        images.extend(pipe(**batch_options).images)
-
-    image_paths = [save_image(img) for img in images]
-    return image_paths, seed
+client = Client(MODEL_ID)
 
 examples = [
     "a cat eating a piece of cheese",
@@ -122,7 +37,7 @@ with gr.Blocks(css=css) as demo:
                 container=False,
             )
             run_button = gr.Button("Run", scale=0)
-        result = gr.Gallery(label="Result", columns=1, show_label=False) 
+        result = gr.Gallery(label="Result", columns=1, show_label=False)
     with gr.Accordion("Advanced options", open=False):
         num_images = gr.Slider(
             label="Number of Images",
@@ -144,7 +59,7 @@ with gr.Blocks(css=css) as demo:
         seed = gr.Slider(
             label="Seed",
             minimum=0,
-            maximum=MAX_SEED,
+            maximum=np.iinfo(np.int32).max,
             step=1,
             value=0,
         )
@@ -193,6 +108,36 @@ with gr.Blocks(css=css) as demo:
         api_name=False,
     )
 
+    def generate(
+        prompt,
+        negative_prompt,
+        use_negative_prompt,
+        seed,
+        width,
+        height,
+        guidance_scale,
+        num_inference_steps,
+        randomize_seed,
+        num_images,
+    ):
+        results = []
+        for _ in range(num_images):
+            response = client.predict(
+                prompt=prompt,
+                negative_prompt=negative_prompt if use_negative_prompt else "",
+                use_negative_prompt=use_negative_prompt,
+                seed=seed,
+                width=width,
+                height=height,
+                guidance_scale=guidance_scale,
+                num_inference_steps=num_inference_steps,
+                randomize_seed=randomize_seed,
+                use_resolution_binning=True,
+                api_name="/run"
+            )
+            results.append(response[0])  # Assuming response contains image path or URL
+        return results, seed
+
     gr.on(
         triggers=[
             prompt.submit,
@@ -217,4 +162,4 @@ with gr.Blocks(css=css) as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.queue(max_size=20).launch()