Spaces:
Sleeping
Sleeping
File size: 3,193 Bytes
184bec9 17633c5 184bec9 17633c5 184bec9 17633c5 184bec9 17633c5 184bec9 17633c5 184bec9 17633c5 8b06ac5 17633c5 8b06ac5 17633c5 8b06ac5 17633c5 8b06ac5 17633c5 184bec9 8b06ac5 |
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 67 68 69 70 71 72 |
import os
import gradio as gr
import asyncio
from langchain_core.prompts import PromptTemplate
from langchain_community.document_loaders import PyPDFLoader
from langchain_google_genai import ChatGoogleGenerativeAI
import google.generativeai as genai
from langchain.chains.question_answering import load_qa_chain
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
# Gemini PDF QA System
async def initialize_gemini(file_path, question):
genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
model = ChatGoogleGenerativeAI(model="gemini-pro", temperature=0.3)
prompt_template = """Answer the question as precise as possible using the provided context. If the answer is
not contained in the context, say "answer not available in context" \n\n
Context: \n {context}?\n
Question: \n {question} \n
Answer:
"""
prompt = PromptTemplate(template=prompt_template, input_variables=["context", "question"])
if os.path.exists(file_path):
pdf_loader = PyPDFLoader(file_path)
pages = pdf_loader.load_and_split()
context = "\n".join(str(page.page_content) for page in pages[:30])
stuff_chain = load_qa_chain(model, chain_type="stuff", prompt=prompt)
stuff_answer = await stuff_chain.acall({"input_documents": pages, "question": question, "context": context}, return_only_outputs=True)
return stuff_answer['output_text']
else:
return "Error: Unable to process the document. Please ensure the PDF file is valid."
# Mistral Text Completion
class MistralModel:
def __init__(self):
self.model_path = "nvidia/Mistral-NeMo-Minitron-8B-Base"
self.tokenizer = AutoTokenizer.from_pretrained(self.model_path)
self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
self.dtype = torch.bfloat16
self.model = AutoModelForCausalLM.from_pretrained(self.model_path, torch_dtype=self.dtype, device_map=self.device)
def generate_text(self, prompt, max_length=50):
inputs = self.tokenizer.encode(prompt, return_tensors='pt').to(self.model.device)
outputs = self.model.generate(inputs, max_length=max_length)
return self.tokenizer.decode(outputs[0])
mistral_model = MistralModel()
# Combined function for both models
async def process_input(file, question):
gemini_answer = await initialize_gemini(file.name, question)
mistral_answer = mistral_model.generate_text(question)
return gemini_answer, mistral_answer
# Gradio Interface
with gr.Blocks() as demo:
gr.Markdown("# PDF Question Answering and Text Completion System")
input_file = gr.File(label="Upload PDF File")
input_question = gr.Textbox(label="Ask a question or provide a prompt")
process_button = gr.Button("Process")
output_text_gemini = gr.Textbox(label="Answer - Gemini")
output_text_mistral = gr.Textbox(label="Answer - Mistral")
process_button.click(
fn=process_input,
inputs=[input_file, input_question],
outputs=[output_text_gemini, output_text_mistral]
)
demo.launch() |