app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "stockmark/stockmark-13b"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")

def generate_text(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=50)
    text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return text

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