app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load your SEA-LION model
tokenizer = AutoTokenizer.from_pretrained("aisingapore/sea-lion-7b-instruct")
model = AutoModelForCausalLM.from_pretrained("aisingapore/sea-lion-7b-instruct")

def generate_response(prompt):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

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

# Serve API request
def api_handler(data):
    return {"response": generate_response(data['input'])}

iface.launch(share=True, inline=True)

# Expose a POST API route using Gradio's internal methods
iface.api_routes = {
    "/generate": {"POST": api_handler}
}