inference.py

from fastai.text.all import load_learner

def chat_with_model():
    learn = load_learner('model.pkl')
    context = "Hey"
    while True:
        print(f"You: {context}")
        generated = learn.predict(context, 200, temperature=0.9)
        print(f"Bot: {generated}\n")

        user_input = input("You: ")
        if user_input.strip().lower() in ["quit", "exit", "stop"]:
            break
        context = user_input

if __name__ == "__main__":
    chat_with_model()