app.py

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

model_name = "m-a-p/MegaBeam-Mistral-7B"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    trust_remote_code=True,
    device_map="auto"
)

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

def chat(prompt):
    output = pipe(prompt, max_new_tokens=512, temperature=0.7)
    return output[0]['generated_text']

iface = gr.Interface(fn=chat, inputs="text", outputs="text", title="MegaBeam Chat 512K")
iface.launch()