Create app.py

app.py

@@ -0,0 +1,85 @@
+# Imports
+import gradio as gr
+import random
+import spaces
+import torch
+import numpy
+import uuid
+import json
+import os
+from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
+from PIL import Image
+
+# Pre-Initialize
+DEVICE = "auto"
+if DEVICE == "auto":
+    DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+print(f"[SYSTEM] | Using {DEVICE} type compute device.")
+
+# Variables
+
+MAX_SEED = 9007199254740991
+DEFAULT_INPUT = ""
+DEFAULT_NEGATIVE_INPUT = ""
+DEFAULT_HEIGHT = 1024
+DEFAULT_WIDTH = 1024
+
+REPO = "sd-community/sdxl-flash"
+REPO_WEIGHT = "ehristoforu/dalle-3-xl-v2"
+WEIGHT = "dalle-3-xl-lora-v2.safetensors"
+ADAPTER = "dalle"
+
+model = StableDiffusionXLPipeline.from_pretrained(REPO, torch_dtype=torch.float16, use_safetensors=True, add_watermarker=False)
+model.scheduler = EulerAncestralDiscreteScheduler.from_config(model.scheduler.config)
+model.load_lora_weights(REPO_WEIGHT, weight_name=WEIGHT, adapter_name=ADAPTER)
+model.set_adapters(ADAPTER, adapter_weights=[0.7])
+model.to(DEVICE)
+
+def get_seed(seed):
+    seed = seed.strip()
+    if seed.isdigit():
+        return int(seed)
+    else:
+        return random.randint(0, MAX_SEED)
+
+@spaces.GPU(duration=30)
+def generate(input=DEFAULT_INPUT, negative_input=DEFAULT_NEGATIVE_INPUT, height=DEFAULT_HEIGHT, width=DEFAULT_WIDTH, steps=1, guidance=0, seed=None):
+    
+    print(input, negative_input, height, width, steps, guidance, seed)
+
+    pipe.to(DEVICE)
+    seed = int(randomize_seed_fn(seed, randomize_seed))
+
+    parameters  = {
+        "prompt": prompt,
+        "negative_prompt": negative_prompt,
+        "height": height,
+        "width": width,
+        "num_inference_steps": steps,
+        "guidance_scale": guidance_scale,
+        "generator": torch.Generator().manual_seed(get_seed(seed)),
+        "use_resolution_binning": True,
+        "output_type":"pil",
+    }
+    
+    images = pipe(**parameters).images
+    image_paths = [save_image(img) for img in images]
+    return image_paths
+
+with gr.Blocks() as main:
+    with gr.Column():
+        input = gr.Textbox(lines=1, value=DEFAULT_INPUT, label="Input")
+        negative_input = gr.Textbox(lines=1, value=DEFAULT_NEGATIVE_INPUT, label="Input Negative")
+        height = gr.Slider(minimum=1, maximum=2160, step=1, value=DEFAULT_HEIGHT, label="Height")
+        width = gr.Slider(minimum=1, maximum=2160, step=1, value=DEFAULT_WIDTH, label="Width")
+        steps = gr.Slider(minimum=0, maximum=100, step=1, value=1, label="Steps")
+        guidance = gr.Slider(minimum=0, maximum=100, step=0.001, value=0, label = "Guidance")
+        seed = gr.Textbox(lines=1, value="", label="Seed (Blank for random)")
+        submit = gr.Button("▶")
+
+    with gr.Column():
+        image = gr.Image(label="Image")
+            
+    submit.click(generate, inputs=[input, negative_input, height, width, steps, guidance, seed], outputs=[image])
+
+main.launch(show_api=True)