Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,112 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
from huggingface_hub import InferenceClient
|
| 3 |
+
from config import HUGGINGFACE_API_KEY # Import your API key from a separate config file
|
| 4 |
+
from PIL import Image
|
| 5 |
+
import requests
|
| 6 |
+
from io import BytesIO
|
| 7 |
+
|
| 8 |
+
# Streamlit App Configuration
|
| 9 |
+
st.set_page_config(page_title="Llama-3.2 Demo App", page_icon="🤖", layout="wide")
|
| 10 |
+
st.title("🖼️ Llama-3.2-90B-Vision-Instruct Demo App")
|
| 11 |
+
st.markdown("<p style='text-align: center; font-size: 18px; color: #555;'>Enter an image URL and get a description</p>", unsafe_allow_html=True)
|
| 12 |
+
|
| 13 |
+
# User Inputs with placeholder
|
| 14 |
+
image_url = st.text_input("Enter Image URL", value="", placeholder="Paste image URL here...", max_chars=400)
|
| 15 |
+
user_prompt = st.text_input("Enter your prompt", value="Describe this image in a paragraph", placeholder="e.g., What is shown in the image?")
|
| 16 |
+
|
| 17 |
+
# Function to display the image from URL with height limit based on its actual size
|
| 18 |
+
def show_image_from_url(image_url, max_height=200):
|
| 19 |
+
try:
|
| 20 |
+
response = requests.get(image_url)
|
| 21 |
+
img = Image.open(BytesIO(response.content))
|
| 22 |
+
|
| 23 |
+
# Get the original image size
|
| 24 |
+
img_width, img_height = img.size
|
| 25 |
+
|
| 26 |
+
# Calculate the new height and width based on the max height while maintaining the aspect ratio
|
| 27 |
+
if img_height > max_height:
|
| 28 |
+
aspect_ratio = img_width / img_height
|
| 29 |
+
new_height = max_height
|
| 30 |
+
new_width = int(new_height * aspect_ratio)
|
| 31 |
+
img_resized = img.resize((new_width, new_height))
|
| 32 |
+
else:
|
| 33 |
+
img_resized = img # No resizing needed if the image is smaller than the max height
|
| 34 |
+
|
| 35 |
+
# Center the image and display it
|
| 36 |
+
st.image(img_resized, caption=f"Source: {image_url}", use_container_width=True)
|
| 37 |
+
|
| 38 |
+
except Exception as e:
|
| 39 |
+
st.error(f"❌ Unable to load image. Error: {e}")
|
| 40 |
+
|
| 41 |
+
# Process user input
|
| 42 |
+
if st.button("Get Description", key="get_description"):
|
| 43 |
+
if image_url and user_prompt:
|
| 44 |
+
try:
|
| 45 |
+
# Show the image with dynamic resizing based on the image size
|
| 46 |
+
show_image_from_url(image_url, max_height=600)
|
| 47 |
+
|
| 48 |
+
# Initialize the InferenceClient
|
| 49 |
+
client = InferenceClient(api_key=HUGGINGFACE_API_KEY)
|
| 50 |
+
|
| 51 |
+
# Define messages for the model
|
| 52 |
+
messages = [
|
| 53 |
+
{
|
| 54 |
+
"role": "user",
|
| 55 |
+
"content": [
|
| 56 |
+
{"type": "text", "text": user_prompt},
|
| 57 |
+
{"type": "image_url", "image_url": {"url": image_url}}
|
| 58 |
+
]
|
| 59 |
+
}
|
| 60 |
+
]
|
| 61 |
+
|
| 62 |
+
# Call the model
|
| 63 |
+
completion = client.chat.completions.create(
|
| 64 |
+
model="meta-llama/Llama-3.2-11B-Vision-Instruct",
|
| 65 |
+
messages=messages,
|
| 66 |
+
max_tokens=500
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
# Extract JSON response
|
| 70 |
+
model_response = completion.choices[0].message
|
| 71 |
+
|
| 72 |
+
# Display the result in a clean and simple format
|
| 73 |
+
st.subheader("📝 Model Response")
|
| 74 |
+
|
| 75 |
+
# Display Content
|
| 76 |
+
st.markdown(f"**Description**: {model_response.get('content', 'No description available')}")
|
| 77 |
+
|
| 78 |
+
except Exception as e:
|
| 79 |
+
st.error(f"❌ An error occurred: {e}")
|
| 80 |
+
else:
|
| 81 |
+
st.warning("⚠️ Please enter an image URL and a prompt.")
|
| 82 |
+
|
| 83 |
+
# Clean UI Enhancements
|
| 84 |
+
st.markdown("""
|
| 85 |
+
<style>
|
| 86 |
+
.stButton>button {
|
| 87 |
+
background-color: #0072BB;
|
| 88 |
+
color: white;
|
| 89 |
+
font-size: 16px;
|
| 90 |
+
border-radius: 10px;
|
| 91 |
+
padding: 10px 20px;
|
| 92 |
+
font-weight: bold;
|
| 93 |
+
transition: background-color 0.3s;
|
| 94 |
+
}
|
| 95 |
+
.stButton>button:hover {
|
| 96 |
+
background-color: #005f8a;
|
| 97 |
+
}
|
| 98 |
+
|
| 99 |
+
.stTextInput>div>div>input {
|
| 100 |
+
padding: 10px;
|
| 101 |
+
font-size: 16px;
|
| 102 |
+
border-radius: 10px;
|
| 103 |
+
}
|
| 104 |
+
|
| 105 |
+
/* Center the image */
|
| 106 |
+
.stImage {
|
| 107 |
+
display: block;
|
| 108 |
+
margin-left: auto;
|
| 109 |
+
margin-right: auto;
|
| 110 |
+
}
|
| 111 |
+
</style>
|
| 112 |
+
""", unsafe_allow_html=True)
|