app.py

import gradio as gr
import peft
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the model and config when the script starts
config = PeftConfig.from_pretrained("PhantHive/llama-2-7b-momo")
model = AutoModelForCausalLM.from_pretrained("NousResearch/Llama-2-7b-chat-hf")
model = PeftModel.from_pretrained(model, "PhantHive/llama-2-7b-momo")

# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-chat-hf")

def greet(text):
    batch = tokenizer(f"'{text}' ->: ", return_tensors='pt')

    with torch.cuda.amp.autocast():
        output_tokens = model.generate(**batch, max_new_tokens=100)

    return tokenizer.decode(output_tokens[0], skip_special_tokens=True)


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