import streamlit as st from transformers import pipeline import torch # Import torch for GPU availability check import json # Define Agent class class Agent: def __init__(self, task, model_name): self.task = task self.model = pipeline(task, model=model_name, device=0 if torch.cuda.is_available() else -1) def run(self, input_data): return self.model(input_data) # Define MoAManager class class MoAManager: def __init__(self): self.agents = {} def add_agent(self, task, model_name): self.agents[task] = Agent(task, model_name) def run_task(self, task, input_data): if task in self.agents: return self.agents[task].run(input_data) else: return "No agent available for this task." # Initialize MoA Manager and add agents moa = MoAManager() moa.add_agent("summarization", "sshleifer/distilbart-cnn-6-6") moa.add_agent("question-answering", "distilbert-base-uncased-distilled-squad") moa.add_agent("text-generation", "gpt2") # Streamlit Interface st.title("Advanced Mixture of Agents (MoA) Demo") st.sidebar.header("Configuration") task = st.sidebar.selectbox("Choose a task", ["summarization", "question-answering", "text-generation"]) st.markdown(""" """, unsafe_allow_html=True) input_text = st.text_area("Input Text", height=200) if task == "question-answering": question = st.text_input("Question", "") if st.button("Run Task"): st.markdown('

Result:

', unsafe_allow_html=True) with st.spinner("Processing..."): if task == "summarization": result = moa.run_task(task, input_text) if result: st.write(result[0]['summary_text']) else: st.error("Error processing summarization task.") elif task == "question-answering": input_data = {"question": question, "context": input_text} result = moa.run_task(task, input_data) formatted_result = json.dumps(result, indent=4) st.code(formatted_result, language='json') elif task == "text-generation": result = moa.run_task(task, input_text) if result: st.write(result[0]['generated_text']) else: st.error("Error processing text-generation task.") # Load data examples st.sidebar.header("Example Data") example_task = st.sidebar.selectbox("Choose example task", ["summarization", "question-answering", "text-generation"]) if example_task == "summarization": st.sidebar.markdown(""" **Example Input:** ``` Artificial intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. The term may also be applied to any machine that exhibits traits associated with a human mind such as learning and problem-solving. AI is continuously evolving to benefit many different industries. Machines are wired using a cross-disciplinary approach based on mathematics, computer science, linguistics, psychology, and more. John McCarthy, an American computer scientist, coined the term 'artificial intelligence' in 1956 at the Dartmouth Conference, where the discipline was born. ``` """) elif example_task == "question-answering": st.sidebar.markdown(""" **Example Context:** ``` Artificial intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. The term may also be applied to any machine that exhibits traits associated with a human mind such as learning and problem-solving. AI is continuously evolving to benefit many different industries. Machines are wired using a cross-disciplinary approach based on mathematics, computer science, linguistics, psychology, and more. John McCarthy, an American computer scientist, coined the term 'artificial intelligence' in 1956 at the Dartmouth Conference, where the discipline was born. ``` **Example Question:** ``` Who coined the term 'artificial intelligence' and when? ``` """) elif example_task == "text-generation": st.sidebar.markdown(""" **Example Input:** ``` Once upon a time, ``` """)