Update app.py

app.py

@@ -4,30 +4,42 @@ import random
 import torch
 from openvino.runtime import Core
 from PIL import Image
-import numpy as np
+from huggingface_hub import hf_hub_download
+import os
 
-print("Initializing OpenVINO... Detecting device (CPU/GPU/VPU)")
+print("🔧 Initializing OpenVINO Runtime...")
 
-# --------- Detect execution provider (GPU if available, else CPU) ---------
+# --------- Detect OpenVINO Device ---------
 core = Core()
 available_devices = core.available_devices
-device = "CPU"  # Default to CPU, can switch to GPU or VPU depending on hardware
+device = "CPU"
 
 if "GPU" in available_devices:
     device = "GPU"
 elif "MYRIAD" in available_devices:
-    device = "MYRIAD"  # VPU
+    device = "MYRIAD"
 elif "HDDL" in available_devices:
-    device = "HDDL"  # For Intel Vision Accelerator
+    device = "HDDL"
+
+print(f"✅ Using device: {device}")
+
+# --------- Download model from HF Hub ---------
+model_repo = "HARRY07979/sd-v1-5-openvino"
+model_filename = "stable-diffusion-v1-5.xml"  # you may need to adjust based on actual file
+
+model_path_xml = hf_hub_download(repo_id=model_repo, filename=model_filename)
+model_path_bin = model_path_xml.replace(".xml", ".bin")
+
+print("📦 Model files downloaded")
 
-print(f"Using device: {device}")
+# --------- Compile OpenVINO model ---------
+compiled_model = core.compile_model(model_path_xml, device)
 
-# --------- Load OpenVINO model ---------
-model_xml = "path_to_model/sd-v1-5-openvino.xml"
-model_bin = "path_to_model/sd-v1-5-openvino.bin"
-compiled_model = core.compile_model(model_xml, device_name=device)
+# Get input and output tensor names
+input_layer = compiled_model.input(0)
+output_layer = compiled_model.output(0)
 
-# --------- Dummy NSFW Detection (placeholder) ---------
+# --------- Dummy NSFW filter ---------
 def detect_nsfw(text: str):
     banned = ["nude", "sex", "porn", "xxx", "nsfw"]
     for word in banned:
@@ -35,13 +47,11 @@ def detect_nsfw(text: str):
             return "NSFW"
     return "SFW"
 
-# --------- Preprocessing function ---------
-def preprocess_input(prompt, width, height):
-    # Apply the necessary preprocessing for the prompt, image size, etc.
-    # Assuming here that we're converting prompt to latent space (placeholder)
-    # You would normally use a tokenizer for Stable Diffusion prompts
-    input_data = np.random.rand(1, 3, height, width).astype(np.float32)  # Dummy data for input
-    return input_data
+# --------- Dummy Preprocessing (you should integrate tokenizer/text encoder for real use) ---------
+def preprocess_input(prompt, width, height, seed):
+    # Just generate dummy latent vector
+    np.random.seed(seed)
+    return np.random.rand(1, 3, height, width).astype(np.float32)
 
 # --------- Inference function ---------
 def infer(
@@ -56,31 +66,25 @@ def infer(
     progress=gr.Progress(track_tqdm=True),
 ):
     if detect_nsfw(prompt) == "NSFW":
-        raise ValueError("Prompt contains NSFW content. Please use a safe prompt.")
+        raise gr.Error("⚠️ Prompt contains NSFW content. Please use a safe prompt.")
 
     if randomize_seed:
         seed = random.randint(0, np.iinfo(np.int32).max)
 
-    # Adjust width and height to be divisible by 8
     width = (width // 8) * 8
     height = (height // 8) * 8
 
-    # Preprocess the input (convert to latent space, etc.)
-    input_data = preprocess_input(prompt, width, height)
+    # Prepare dummy input for now
+    input_data = preprocess_input(prompt, width, height, seed)
 
-    # Run inference on OpenVINO
-    input_tensor = compiled_model.input(0)  # Get input layer
-    output_tensor = compiled_model.output(0)  # Get output layer
+    # Run OpenVINO inference
+    result = compiled_model([input_data])[output_layer]
 
-    # Inference
-    result = compiled_model([input_data])[output_tensor]
-
-    # Post-process the result (convert to image)
-    output_image = np.random.rand(3, width, height)  # Dummy image for now
-    output_image = np.transpose(output_image, (1, 2, 0))  # Convert from (C, H, W) to (H, W, C)
+    # Postprocess dummy result
+    output_image = np.clip(result[0].transpose(1, 2, 0), 0, 1)
     output_image = (output_image * 255).astype(np.uint8)
     pil_image = Image.fromarray(output_image)
-    
+
     return pil_image, seed
 
 # --------- UI ---------
@@ -90,80 +94,29 @@ examples = [
     "A bowl of ramen in anime style",
 ]
 
-css = """
-#col-container {
-    margin: 0 auto;
-    max-width: 640px;
-}
-"""
-
-with gr.Blocks(css=css) as demo:
+with gr.Blocks(css="#col-container { max-width: 640px; margin: auto; }") as demo:
     with gr.Column(elem_id="col-container"):
-        gr.Markdown("# SD 1.5 Turbo Lite OpenVINO 🚀")
+        gr.Markdown("## 🧠 SD 1.5 OpenVINO via HARRY07979 on HuggingFace")
 
         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, variant="primary")
+            prompt = gr.Text(label="Prompt", placeholder="Enter your prompt", show_label=False)
+            run_button = gr.Button("Generate", variant="primary")
 
-        result = gr.Image(label="Result", show_label=False)
+        result = gr.Image(label="Output", show_label=False)
 
         with gr.Accordion("Advanced Settings", open=False):
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=True,
-            )
-
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=np.iinfo(np.int32).max,
-                step=1,
-                value=0,
-            )
-
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+            negative_prompt = gr.Text(label="Negative Prompt", placeholder="e.g. low quality, blurry")
+
+            seed = gr.Slider(label="Seed", minimum=0, maximum=np.iinfo(np.int32).max, step=1, value=0)
+            randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
 
             with gr.Row():
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=1024,
-                    step=8,
-                    value=512,
-                )
-
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=1024,
-                    step=8,
-                    value=512,
-                )
+                width = gr.Slider(label="Width", minimum=256, maximum=1024, step=8, value=512)
+                height = gr.Slider(label="Height", minimum=256, maximum=1024, step=8, value=512)
 
             with gr.Row():
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=15.0,
-                    step=0.1,
-                    value=5.0,
-                )
-
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=30,
-                    step=1,
-                    value=15,
-                )
+                guidance_scale = gr.Slider(label="Guidance Scale", minimum=0.0, maximum=15.0, step=0.1, value=7.5)
+                num_inference_steps = gr.Slider(label="Steps", minimum=1, maximum=50, step=1, value=25)
 
         gr.Examples(examples=examples, inputs=[prompt])