Spaces:
Runtime error
Runtime error
Upload 2 files
Browse files- app.py +81 -0
- requirements.txt +9 -0
app.py
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from dotenv import load_dotenv
|
| 2 |
+
import os
|
| 3 |
+
import streamlit as st
|
| 4 |
+
from PIL import Image
|
| 5 |
+
from transformers import ViltProcessor, ViltForQuestionAnswering
|
| 6 |
+
from langchain.prompts import PromptTemplate
|
| 7 |
+
from langchain.chains import LLMChain
|
| 8 |
+
from streamlit_extras.add_vertical_space import add_vertical_space
|
| 9 |
+
from langchain.llms import OpenAI
|
| 10 |
+
|
| 11 |
+
load_dotenv()
|
| 12 |
+
|
| 13 |
+
processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
| 14 |
+
model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
|
| 15 |
+
|
| 16 |
+
llm = OpenAI(temperature=0.2)
|
| 17 |
+
prompt = PromptTemplate(
|
| 18 |
+
input_variables=["question", "elements"],
|
| 19 |
+
template="""Please generate a structured response using the following information:
|
| 20 |
+
\n\n
|
| 21 |
+
#Question: {question}
|
| 22 |
+
#Response: {elements}
|
| 23 |
+
\n\n
|
| 24 |
+
Your structured response:""",
|
| 25 |
+
)
|
| 26 |
+
|
| 27 |
+
def process_query(image, query):
|
| 28 |
+
encoding = processor(image, query, return_tensors="pt")
|
| 29 |
+
outputs = model(**encoding)
|
| 30 |
+
logits = outputs.logits
|
| 31 |
+
idx = logits.argmax(-1).item()
|
| 32 |
+
chain = LLMChain(llm=llm, prompt=prompt)
|
| 33 |
+
response = chain.run(question=query, elements=model.config.id2label[idx])
|
| 34 |
+
return response
|
| 35 |
+
|
| 36 |
+
st.set_page_config(page_title="Insightly")
|
| 37 |
+
# Sidebar contents
|
| 38 |
+
with st.sidebar:
|
| 39 |
+
st.sidebar.image("/home/oem/Downloads/insightly_wbg.png", use_column_width=True)
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
load_dotenv()
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
def main():
|
| 48 |
+
st.title("Chat With Images 🖼️💬")
|
| 49 |
+
|
| 50 |
+
uploaded_file = st.file_uploader('Upload your Image', type=['png', 'jpeg', 'jpg'])
|
| 51 |
+
|
| 52 |
+
if uploaded_file is not None:
|
| 53 |
+
|
| 54 |
+
image = Image.open(uploaded_file)
|
| 55 |
+
st.image(image, caption='Uploaded Image.', width=300)
|
| 56 |
+
|
| 57 |
+
cancel_button = st.button('Remove this image')
|
| 58 |
+
if cancel_button:
|
| 59 |
+
st.markdown(
|
| 60 |
+
"""<style>
|
| 61 |
+
.css-2trqyj:focus,
|
| 62 |
+
.css-2trqyj:active,
|
| 63 |
+
.css-2trqyj:hover {
|
| 64 |
+
outline: 2px solid red;
|
| 65 |
+
outline-offset: 2px;
|
| 66 |
+
}
|
| 67 |
+
</style>""",
|
| 68 |
+
unsafe_allow_html=True
|
| 69 |
+
)
|
| 70 |
+
query = st.text_input('Type your question here')
|
| 71 |
+
|
| 72 |
+
if query:
|
| 73 |
+
with st.spinner('Processing...'):
|
| 74 |
+
answer = process_query(image, query)
|
| 75 |
+
st.write(answer)
|
| 76 |
+
|
| 77 |
+
if cancel_button:
|
| 78 |
+
st.stop()
|
| 79 |
+
|
| 80 |
+
if __name__ == "__main__":
|
| 81 |
+
main()
|
requirements.txt
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
streamlit
|
| 2 |
+
python-dotenv
|
| 3 |
+
Pillow
|
| 4 |
+
transformers
|
| 5 |
+
streamlit-extras
|
| 6 |
+
langchain
|
| 7 |
+
torch
|
| 8 |
+
openai
|
| 9 |
+
tokenizers
|