app.py

import gradio as gr

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer, pipeline
from threading import Thread

model_id = "rasyosef/Llama-3.2-400M-Amharic-Instruct-Poems-Stories-Wikipedia"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float32,
    # device_map="cuda" if torch.cuda.is_available() else "cpu"
  )

llama3_am = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    eos_token_id=tokenizer.eos_token_id,
    # device_map="cuda" if torch.cuda.is_available() else "cpu"
  )

# Function that accepts a prompt and generates text 
def generate(message, chat_history, max_new_tokens=64):

  history = []

  for sent, received in chat_history:
    history.append({"role": "user", "content": sent})
    history.append({"role": "assistant", "content": received})

  history.append({"role": "user", "content": message})

  if len(tokenizer.apply_chat_template(history)) > 512:
    yield "chat history is too long"
  else:
    # Streamer
    streamer = TextIteratorStreamer(tokenizer=tokenizer, skip_prompt=True, skip_special_tokens=True, timeout=300.0)
    thread = Thread(
        target=llama3_am,
        kwargs={
            "text_inputs":history,
            "max_new_tokens":max_new_tokens,
            "repetition_penalty":1.1,
            "streamer":streamer
            }
        )
    thread.start()

    generated_text = ""
    for word in streamer:
      generated_text += word
      response = generated_text.strip()

      yield response

# Chat interface with gradio
with gr.Blocks() as demo:
  gr.Markdown("""
  # Llama 3.2 400M Amharic Chatbot Demo
  """)

  tokens_slider = gr.Slider(8, 256, value=64, label="Maximum new tokens", info="A larger `max_new_tokens` parameter value gives you longer text responses but at the cost of a slower response time.")

  chatbot = gr.ChatInterface(
    chatbot=gr.Chatbot(height=400),
    fn=generate,
    additional_inputs=[tokens_slider],
    stop_btn=None,
    cache_examples=False,
    examples=[
        ["ሰላም"],
        ["ሰላም፣ እንዴት ነህ?"],
        ["አንተ ማነህ?"],
        ["ግጥም ፃፍልኝ"],
        ["ስለ ይቅርታ ግጥም ጻፍልኝ"],
        ["አንድ ተረት አጫውተኝ"],
        ["ስለ ጅብና አንበሳ ተረት ንገረኝ"],
        ["ቀልድ ንገረኝ"],
        ["ስለ ስራ አጥነት አንድ ቀልድ ንገረኝ"],
        ["ዳግማዊ ቴዎድሮስ ማን ነው?"],
        ["ዳግማዊ ምኒልክ ማን ነው?"],
        ["ስለ አዲስ አበባ ዩኒቨርስቲ ጥቂት እውነታዎችን አጫውተኝ"],
        ["ስለ ጃፓን ጥቂት እውነታዎችን ንገረኝ"],
        ["ስለ ማይክሮሶፍት ጥቂት እውነታዎችን ንገረኝ"],
        ["ጉግል ምንድን ነው?"],
        ["ቢትኮይን ምንድን ነው?"],
      ]
  )

demo.queue().launch(debug=True)