Spaces:
Sleeping
Sleeping
| import os | |
| import gradio as gr | |
| import asyncio | |
| from langchain_core.prompts import PromptTemplate | |
| from langchain_community.output_parsers.rail_parser import GuardrailsOutputParser | |
| 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(file_path, question): | |
| genai.configure(api_key=os.getenv("GOOGLE_API_KEY")) | |
| model = genai.GenerativeModel('gemini-pro') | |
| 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({"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." | |
| async def pdf_qa(file, question): | |
| answer = await initialize(file.name, question) | |
| return answer | |
| # Mistral Text Completion | |
| def load_mistral_model(): | |
| model_path = "nvidia/Mistral-NeMo-Minitron-8B-Base" | |
| tokenizer = AutoTokenizer.from_pretrained(model_path) | |
| device = 'cuda' if torch.cuda.is_available() else 'cpu' | |
| dtype = torch.bfloat16 | |
| model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=dtype, device_map=device) | |
| return tokenizer, model | |
| def generate_text(prompt, max_length=50): | |
| tokenizer, model = load_mistral_model() | |
| inputs = tokenizer.encode(prompt, return_tensors='pt').to(model.device) | |
| outputs = model.generate(inputs, max_length=max_length) | |
| return tokenizer.decode(outputs[0]) | |
| # Gradio Interface | |
| def pdf_qa_wrapper(file, question): | |
| return asyncio.run(pdf_qa(file, question)) | |
| with gr.Blocks() as demo: | |
| gr.Markdown("# Combined PDF QA and Text Completion System") | |
| with gr.Tab("PDF Question Answering"): | |
| input_file = gr.File(label="Upload PDF File") | |
| input_question = gr.Textbox(label="Ask about the document") | |
| output_text_gemini = gr.Textbox(label="Answer - GeminiPro") | |
| pdf_qa_button = gr.Button("Ask Question") | |
| with gr.Tab("Text Completion"): | |
| input_prompt = gr.Textbox(label="Enter prompt for text completion") | |
| output_text_mistral = gr.Textbox(label="Completed Text - Mistral") | |
| complete_text_button = gr.Button("Complete Text") | |
| pdf_qa_button.click(pdf_qa_wrapper, inputs=[input_file, input_question], outputs=output_text_gemini) | |
| complete_text_button.click(generate_text, inputs=input_prompt, outputs=output_text_mistral) | |
| demo.launch() |