import gradio as gr

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

model_id = "rasyosef/llama-3.2-amharic-64k-instruct-beta"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

llama_am = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    pad_token_id=tokenizer.pad_token_id,
    eos_token_id=tokenizer.eos_token_id
  )

# Function that accepts a prompt and generates text using the phi2 pipeline
def generate(message, chat_history, max_new_tokens=256):

  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})
  #print(history)

  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=llama_am,
                    kwargs={
                        "text_inputs":history,
                        "max_new_tokens":max_new_tokens,
                        "temperature":0.2,
                        "do_sample":True,
                        "top_k":4,
                        "top_p":0.8,
                        "repetition_penalty":1.25,
                        "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 Amharic Chatbot Demo

  This chatbot was created using [Llama-3.2-Amharic-Instruct](https://huggingface.co/rasyosef/llama-3.2-amharic-64k-instruct-beta), a finetuned version of my 180 million parameter [Llama 3.2 Amharic](https://huggingface.co/rasyosef/llama-3.2-amharic-64k-1024) transformer model.
  """)

  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=[
        ["የኢትዮጵያ ዋና ከተማ ስም ምንድን ነው?"],
        ["የፈረንሳይ ዋና ከተማ ስም ምንድን ነው?"],
        ["የኢትዮጵያ የመጨረሻው ንጉስ ማን ነበሩ?"],
        ["የአሜሪካ ፕሬዝዳንት ማን ነው?"],
        ["የእስራኤል ጠቅላይ ሚንስትር ማን ነው?"],
        ["ሶስት የአፍሪካ ሃገራት ጥቀስልኝ"],
        ["፫ የአውሮፓ ሃገራት ጥቀስልኝ"],
        ["፫ የእስያ ሃገራት ጥቀስልኝ"],
        ["ሶስት የአሜሪካ መሪዎችን ስም ጥቀስ"],
        ["አምስት የአሜሪካ ከተማዎችን ስም ጥቀስ"],
        #["አልበርት አይንስታይን ማን ነው?"],
      ]
  )

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