File size: 2,168 Bytes
2bf176d baf8eca 1348715 baf8eca cc851d6 baf8eca cc851d6 2bf176d 0e8a17c baf8eca cc851d6 0e8a17c baf8eca 2379c94 baf8eca 2379c94 cc851d6 2379c94 2bda478 2379c94 cc851d6 2379c94 cc851d6 2379c94 cc851d6 2379c94 2bda478 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
import os
import streamlit as st
from PIL import Image
import torch
from transformers import PaliGemmaProcessor, PaliGemmaForConditionalGeneration
# Get Hugging Face API key from environment variables
HF_TOKEN = os.getenv("HF_KEY")
# Ensure API key is available
if not HF_TOKEN:
st.error("❌ Hugging Face API key not found! Set it as 'HF_KEY' in Spaces secrets.")
st.stop()
# Load the model and processor with authentication
@st.cache_resource
def load_model():
model_id = "google/paligemma2-3b-mix-224"
try:
model = PaliGemmaForConditionalGeneration.from_pretrained(
model_id, token=HF_TOKEN, torch_dtype=torch.bfloat16, device_map="auto"
).eval()
processor = PaliGemmaProcessor.from_pretrained(
model_id, token=HF_TOKEN
)
return processor, model
except Exception as e:
st.error(f"❌ Error loading model: {str(e)}")
st.stop()
processor, model = load_model()
# Streamlit UI
st.title("🖼️ Image Understanding with PaliGemma")
uploaded_file = st.file_uploader("Upload an Image", type=["png", "jpg", "jpeg"])
if uploaded_file:
image = Image.open(uploaded_file).convert("RGB")
st.image(image, caption="Uploaded Image", use_container_width=True)
# User input for task selection
task = st.selectbox(
"Select a task:",
["Generate a caption", "Answer a question", "Detect objects", "Generate segmentation"]
)
# User prompt
prompt = st.text_area("Enter a prompt (e.g., 'Describe the image' or 'What objects are present?')")
if st.button("Run"):
if prompt:
inputs = processor(text=prompt, images=image, return_tensors="pt").to(torch.bfloat16).to(model.device)
input_len = inputs["input_ids"].shape[-1] # Get input length
with torch.inference_mode():
generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
generation = generation[0][input_len:] # Remove input tokens from output
answer = processor.decode(generation, skip_special_tokens=True)
st.success(f"✅ Result: {answer}")
|