from transformers import AutoModelForCausalLM, AutoTokenizer
from instruct_pipeline import InstructionTextGenerationPipeline
import torch
tokenizer = AutoTokenizer.from_pretrained("databricks/dolly-v2-3b", padding_side="left")
model = AutoModelForCausalLM.from_pretrained("databricks/dolly-v2-3b", torch_dtype=torch.bfloat16)
                                             
generate_text = InstructionTextGenerationPipeline(model=model, tokenizer=tokenizer)

import gradio as gr

def greet(question):
    return generate_text(question)
    

iface = gr.Interface(fn=greet, inputs="text", outputs="text")
iface.launch()