import os
os.system('pip install -q git+https://github.com/huggingface/transformers.git')
os.system('pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu')
os.system('pip install fitz')
os.system('pip install PyMuPDF')
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
import gradio as gr
import re
import fitz
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-large")
model = AutoModelForSeq2SeqLM.from_pretrained("google/flan-t5-large").to(device)
class GUI:
def preprocess(self,text):
text = text.replace('\n', ' ')
text = re.sub('\s+', ' ', text)
return text
def query_from_list(self,query, options, tok_len):
t5query = f"""Question: "{query}" Context: {options}"""
inputs = tokenizer(t5query, return_tensors="pt").to(device)
outputs = model.generate(**inputs, max_new_tokens=tok_len)
return tokenizer.batch_decode(outputs, skip_special_tokens=True)
def begin(self,pdf,question,start_page=1, end_page=None):
doc = fitz.open(pdf)
total_pages = doc.page_count
if end_page is None:
end_page = total_pages
pdf_text = ""
for i in range(start_page-1, end_page):
text = doc.load_page(i).get_text("text")
text = app.preprocess(text)
pdf_text+=text
# Call the LLM with input data and instruction
input_data=pdf_text
results = app.query_from_list(question, input_data, 30)
return results
app = GUI()
title = "Get answers from your document with questions with Flan-T5"
description = "Results will show up in a few seconds."
article="References
[1] FLAN-T5” Transformers Link
"
css = """.output_image, .input_image {height: 600px !important}"""
iface = gr.Interface(fn=app.begin,
inputs=[gr.File(label="PDF File",file_types=['.pdf']), gr.Textbox(label="Question") ],
outputs = gr.Text(label="Answer Summary"),
title=title,
description=description,
article=article,
css=css,
analytics_enabled = True, enable_queue=True)
iface.launch(inline=False, share=False, debug=False)