Create app.py

app.py

@@ -0,0 +1,66 @@
+import streamlit as st
+from PIL import Image
+from ultralytics import YOLO
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+import torch
+
+# Load YOLO model
+yolo_model = YOLO('runs/detect/makkah-yolo/weights/best.pt')  # Replace with your relative path or uploaded model if needed
+
+# Load ALLaM in 8-bit to reduce memory
+bnb_config = BitsAndBytesConfig(load_in_8bit=True)
+llm_model = AutoModelForCausalLM.from_pretrained(
+    "ALLaM-AI/ALLaM-7B-Instruct-preview",
+    quantization_config=bnb_config,
+    device_map="auto",
+    trust_remote_code=True
+)
+tokenizer = AutoTokenizer.from_pretrained("ALLaM-AI/ALLaM-7B-Instruct-preview", trust_remote_code=True)
+
+# Streamlit App UI
+st.title("🔎🕋 Makkah Landmark Detection + Arabic Cultural Explanation")
+st.write("Upload an image to detect landmarks in Makkah and receive a cultural explanation in Arabic using ALLaM.")
+
+uploaded_file = st.file_uploader("📷 Upload an image...", type=["jpg", "jpeg", "png"])
+
+if uploaded_file:
+    img = Image.open(uploaded_file)
+    st.image(img, caption="Uploaded Image", use_column_width=True)
+
+    temp_image_path = "temp_uploaded_image.jpg"
+    img.save(temp_image_path)
+
+    with st.spinner("🔍 Detecting landmarks..."):
+        results = yolo_model.predict(temp_image_path, conf=0.25, save=False)
+
+    detections = results[0].names
+    boxes = results[0].boxes.cls
+    detected_classes = [detections[int(cls_id)] for cls_id in boxes]
+
+    if detected_classes:
+        landmarks = '، '.join(detected_classes)
+        user_prompt = f"من فضلك، تحدث عن أهمية المعالم التالية في مكة المكرمة: {landmarks}."
+
+        st.write("📜 **Arabic Prompt:**")
+        st.info(user_prompt)
+
+        with st.spinner("🤖 Generating Arabic explanation..."):
+            messages = [{"role": "user", "content": user_prompt}]
+            inputs = tokenizer.apply_chat_template(messages, tokenize=False)
+            inputs = tokenizer(inputs, return_tensors='pt', return_token_type_ids=False)
+            inputs = {k: v.to('cuda') for k, v in inputs.items()}
+
+            response = llm_model.generate(
+                **inputs,
+                max_new_tokens=512,
+                do_sample=True,
+                top_k=50,
+                top_p=0.95,
+                temperature=0.6
+            )
+
+            output_text = tokenizer.batch_decode(response, skip_special_tokens=True)[0]
+            st.write("🕌 **Explanation in Arabic:**")
+            st.success(output_text)
+    else:
+        st.warning("🚫 No landmarks detected in this image.")