✨ | Use SDXL from rupeshs instead

main.py

@@ -1,33 +1,30 @@
 from flask import Flask, request, jsonify
-from diffusers import DDPMScheduler, DiffusionPipeline
-import torch
+from optimum.intel.openvino.modeling_diffusion import OVStableDiffusionXLPipeline
 import base64
 from io import BytesIO
 from PIL import Image
+import os
 
 app = Flask(__name__)
 
-scheduler = DDPMScheduler.from_pretrained("stabilityai/stable-diffusion-2-1-base", subfolder="scheduler")
-pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2-1-base", scheduler=scheduler)
+# Define paths for OpenVINO IR models
+OV_MODEL_DIR = "echarlaix/stable-diffusion-2-1-openvino" # New directory for the LCM model
+
+# --- Model Loading ---
+pipe = None
 
-print("Loading UNet weights...")
-ckpt_path = './models/model.bin'
 try:
-    unet_state_dict = torch.load(ckpt_path)
-    pipe.unet.load_state_dict(unet_state_dict)
-    if torch.cuda.is_available():
-        pipe = pipe.to("cuda")
-        pipe.to(torch.float16)
-        print("UNet weights loaded successfully and model moved to CUDA.")
-    else:
-        print("CUDA is not available. Running on CPU (will be slow).")
-        pipe = pipe.to("cpu")
-        pipe.to(torch.float32)
-except FileNotFoundError:
-    print(f"Error: Model not found at {ckpt_path}. Please ensure the file exists.")
-    exit()
+    pipe = OVStableDiffusionXLPipeline.from_pretrained(
+        "rupeshs/hyper-sd-sdxl-1-step-openvino-int8",
+        ov_config={"CACHE_DIR": ""},
+    )
+
+    print("Compiling OpenVINO pipeline...")
+    pipe.compile() # Compile the pipeline for the target device (CPU by default)
+    print("OpenVINO compiled successfully.")
+
 except Exception as e:
-    print(f"An error occurred while loading UNet weights: {e}")
+    print(f"An error occurred during OpenVINO LCM model loading or conversion: {e}")
     exit()
 
 @app.route('/api/generate', methods=['POST'])
@@ -36,23 +33,28 @@ def generate_image():
     if not prompt:
         return jsonify({"error": "Prompt is required in the request body."}), 400
 
-    print(f"Generating image for prompt: '{prompt}'...")
+    print(f"Generating image for prompt: '{prompt}' using OpenVINO LCM...")
     try:
-        image = pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, timesteps=[999]).images[0]
+        # Crucially, set num_inference_steps=1 for 1-step generation with LCMs
+        # You might even omit guidance_scale for some LCMs, or use a very low value.
+        image = pipe(
+            prompt=prompt,
+            width=768,
+            height=768,
+            num_inference_steps=1,
+            guidance_scale=1.0,
+        ).images[0]
         image = image.resize((128, 128), Image.LANCZOS)
 
-        # Convert image to base64
         buffered = BytesIO()
-        image.save(buffered, format="JPEG") # You can choose JPEG for smaller size, but PNG is lossless
+        image.save(buffered, format="PNG")
         img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
 
         return jsonify({"image": img_str})
 
     except Exception as e:
-        print(f"Error during image generation: {e}")
+        print(f"Error during image generation with OpenVINO LCM: {e}")
         return jsonify({"error": "An error occurred during image generation."}), 500
 
 if __name__ == '__main__':
-    # Make sure to run this with 'python app.py'
-    # It will be accessible at http://127.0.0.1:5000/
-    app.run(host='127.0.0.1', port=5000, debug=True) # debug=True is good for development
+    app.run(host='127.0.0.1', port=5000, debug=True)