app.py

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

# Load model and tokenizer
model_name = "Spestly/Athena-1-1.5B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float32, low_cpu_mem_usage=True)

# Set to evaluation mode
model.eval()

def generate_response(message, history):
    instruction = (
        "You are an LLM called Athena. You are finetuned by Aayan Mishra. You are NOT trained by Anthropic. "
        "You are a Qwen 2.5 fine-tune. Your purpose is the help the user accomplish their request to the best of your abilities. "
        "Below is an instruction that describes a task. Answer it clearly and concisely.\n\n"
        f"### Instruction:\n{message}\n\n### Response:"
    )
    
    inputs = tokenizer(instruction, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=1000,
            num_return_sequences=1,
            temperature=0.7,
            top_p=0.9,
            do_sample=True
        )
    
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    response = response.split("### Response:")[-1].strip()
    
    return response

iface = gr.ChatInterface(
    generate_response,
    chatbot=gr.Chatbot(height=600, type="messages"),
    textbox=gr.Textbox(placeholder="Type your message here...", container=False, scale=7),
    title="Athena-1",
    description="Chat with Athena-1",
    theme="soft",
    examples=[
        "Can you give me a good salsa recipe?",
        "Write an engaging two-line horror story.",
        "What is the capital of Australia?",
    ],
    type="messages"
)

iface.launch()