Upload 2 files

app.py

@@ -112,23 +112,13 @@ iface = gr.Interface(
     title="Image Question Answering with Phi-3 and SigLIP (CPU)",
     description="Ask questions about an image and get answers powered by Phi-3 (llama.cpp) and SigLIP.",
     examples=[
-        ["example_image_1.jpg", "What color is the car?"],
-        ["example_image_2.jpg", "How many people are in the image?"],
-        ["example_image_3.jpg", "What is the person doing?"]
+        ["cat_0006.png", "Create a interesting story about this image?"],
+        ["bird_0004.png", "Can you describe this image?"],
+        ["truck_0003.png", "Elaborate the setting of the image"],
+        ["ship_0007.png", "Explain the purpose of image"]
     ]
 )
 
 # 5. Launch the App
 if __name__ == "__main__":
-    # Create dummy example images
-    if not os.path.exists("example_image_1.jpg"):
-        dummy_image = Image.new("RGB", (224, 224), color="red")
-        dummy_image.save("example_image_1.jpg")
-    if not os.path.exists("example_image_2.jpg"):
-        dummy_image = Image.new("RGB", (224, 224), color="green")
-        dummy_image.save("example_image_2.jpg")
-    if not os.path.exists("example_image_3.jpg"):
-        dummy_image = Image.new("RGB", (224, 224), color="blue")
-        dummy_image.save("example_image_3.jpg")
-
     iface.launch()

requirements.txt

@@ -4,4 +4,5 @@ torchvision
 timm
 Pillow 
 transformers
-llama-cpp-python
+llama-cpp-python
+peft