app.py

import streamlit as st
import pandas as pd
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch
from datasets import load_dataset

# Load model and tokenizer
model_path = "rb757/new_app"
model = AutoModelForSequenceClassification.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)

# Load the dataset
dataset_dict = load_dataset(
    'HUPD/hupd',
    name='sample',
    data_files="https://huggingface.co/datasets/HUPD/hupd/resolve/main/hupd_metadata_2022-02-22.feather",
    train_filing_start_date='2016-01-01',
    train_filing_end_date='2016-01-21',
    val_filing_start_date='2016-01-22',
    val_filing_end_date='2016-01-31',
    trust_remote_code=True
)

# Convert to DataFrame
train_df = pd.DataFrame(dataset_dict['train'])
val_df = pd.DataFrame(dataset_dict['validation'])

# Print columns to verify availability
print("Train set columns:", train_df.columns.tolist())
print("Validation set columns:", val_df.columns.tolist())

# Title and description
st.title("Milestone Patent 🐨")
st.write("Select a patent application to evaluate its patentability.")

# Dropdown for patent numbers
patent_numbers = train_df['patent_number'].unique()
selected_patent = st.selectbox("Select Patent Number", patent_numbers)

# Retrieve abstract and claims
if selected_patent:
    patent_info = train_df[train_df['patent_number'] == selected_patent].iloc[0]
    abstract = patent_info['abstract']
    claims = patent_info['claims']

    # Display the abstract and claims
    st.text_area("Abstract", abstract, height=150)
    st.text_area("Claims", claims, height=150)

    # Submit button
    if st.button("Get Patentability Score"):
        # Prepare the input text
        input_text = f"{abstract} {claims}"
        inputs = tokenizer(input_text, return_tensors="pt", truncation=True, padding=True)

        # Get the model prediction
        with torch.no_grad():
            logits = model(**inputs).logits
            predictions = torch.argmax(logits, dim=-1)

        # Display the patentability score
        decision_labels = ['REJECTED', 'ACCEPTED', 'PENDING', 'CONT-REJECTED', 'CONT-ACCEPTED', 'CONT-PENDING']
        score = decision_labels[predictions.item()]
        st.write(f"Patentability Score: **{score}**")