import os import gradio as gr from text_generation import Client from conversation import get_default_conv_template from transformers import AutoTokenizer DESCRIPTION = """ # Language Models for Taiwanese Culture

✍ī¸ Online Demo â€ĸ 🤗 HF Repo â€ĸ đŸĻ Twitter â€ĸ 📃 [Paper Coming Soon] â€ĸ 👨ī¸ Yen-Ting Lin


Taiwan-LLaMa is a fine-tuned model specifically designed for traditional Chinese applications. It is built upon the LLaMa 2 architecture and includes a pretraining phase with over 5 billion tokens and fine-tuning with over 490k multi-turn conversational data in Traditional Chinese. ## Key Features 1. **Traditional Chinese Support**: The model is fine-tuned to understand and generate text in Traditional Chinese, making it suitable for Taiwanese culture and related applications. 2. **Instruction-Tuned**: Further fine-tuned on conversational data to offer context-aware and instruction-following responses. 3. **Performance on Vicuna Benchmark**: Taiwan-LLaMa's relative performance on Vicuna Benchmark is measured against models like GPT-4 and ChatGPT. It's particularly optimized for Taiwanese culture. 4. **Flexible Customization**: Advanced options for controlling the model's behavior like system prompt, temperature, top-p, and top-k are available in the demo. ## Model Versions Different versions of Taiwan-LLaMa are available: - **Taiwan-LLaMa v1.0 (This demo)**: Optimized for Taiwanese Culture - **Taiwan-LLaMa v0.9**: Partial instruction set - **Taiwan-LLaMa v0.0**: No Traditional Chinese pretraining The models can be accessed from the provided links in the Hugging Face repository. Try out the demo to interact with Taiwan-LLaMa and experience its capabilities in handling Traditional Chinese! """ LICENSE = """ ## Licenses - Code is licensed under Apache 2.0 License. - Models are licensed under the LLAMA 2 Community License. - By using this model, you agree to the terms and conditions specified in the license. - By using this demo, you agree to share your input utterances with us to improve the model. ## Acknowledgements Taiwan-LLaMa project acknowledges the efforts of the [Meta LLaMa team](https://github.com/facebookresearch/llama) and [Vicuna team](https://github.com/lm-sys/FastChat) in democratizing large language models. """ DEFAULT_SYSTEM_PROMPT = "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. You are built by NTU Miulab by Yen-Ting Lin for research purpose." endpoint_url = os.environ.get("ENDPOINT_URL", "http://127.0.0.1:8080") client = Client(endpoint_url, timeout=120) eos_token = "" MAX_MAX_NEW_TOKENS = 1024 DEFAULT_MAX_NEW_TOKENS = 1024 max_prompt_length = 4096 - MAX_MAX_NEW_TOKENS - 10 tokenizer = AutoTokenizer.from_pretrained("yentinglin/Taiwan-LLaMa-v1.0") with gr.Blocks() as demo: gr.Markdown(DESCRIPTION) chatbot = gr.Chatbot() with gr.Row(): msg = gr.Textbox( container=False, show_label=False, placeholder='Type a message...', scale=10, ) submit_button = gr.Button('Submit', variant='primary', scale=1, min_width=0) with gr.Row(): retry_button = gr.Button('🔄 Retry', variant='secondary') undo_button = gr.Button('↩ī¸ Undo', variant='secondary') clear = gr.Button('🗑ī¸ Clear', variant='secondary') saved_input = gr.State() with gr.Accordion(label='Advanced options', open=False): system_prompt = gr.Textbox(label='System prompt', value=DEFAULT_SYSTEM_PROMPT, lines=6) max_new_tokens = gr.Slider( label='Max new tokens', minimum=1, maximum=MAX_MAX_NEW_TOKENS, step=1, value=DEFAULT_MAX_NEW_TOKENS, ) temperature = gr.Slider( label='Temperature', minimum=0.1, maximum=1.0, step=0.1, value=0.7, ) top_p = gr.Slider( label='Top-p (nucleus sampling)', minimum=0.05, maximum=1.0, step=0.05, value=0.9, ) top_k = gr.Slider( label='Top-k', minimum=1, maximum=1000, step=1, value=50, ) def user(user_message, history): return "", history + [[user_message, None]] def bot(history, max_new_tokens, temperature, top_p, top_k, system_prompt): conv = get_default_conv_template("vicuna").copy() roles = {"human": conv.roles[0], "gpt": conv.roles[1]} # map human to USER and gpt to ASSISTANT conv.system = system_prompt for user, bot in history: conv.append_message(roles['human'], user) conv.append_message(roles["gpt"], bot) msg = conv.get_prompt() prompt_tokens = tokenizer.encode(msg) length_of_prompt = len(prompt_tokens) if length_of_prompt > max_prompt_length: msg = tokenizer.decode(prompt_tokens[-max_prompt_length + 1:]) history[-1][1] = "" for response in client.generate_stream( msg, max_new_tokens=max_new_tokens, temperature=temperature, top_p=top_p, top_k=top_k, ): if not response.token.special: character = response.token.text history[-1][1] += character yield history msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then( fn=bot, inputs=[ chatbot, max_new_tokens, temperature, top_p, top_k, system_prompt, ], outputs=chatbot ) submit_button.click( user, [msg, chatbot], [msg, chatbot], queue=False ).then( fn=bot, inputs=[ chatbot, max_new_tokens, temperature, top_p, top_k, system_prompt, ], outputs=chatbot ) def delete_prev_fn( history: list[tuple[str, str]]) -> tuple[list[tuple[str, str]], str]: try: message, _ = history.pop() except IndexError: message = '' return history, message or '' def display_input(message: str, history: list[tuple[str, str]]) -> list[tuple[str, str]]: history.append((message, '')) return history retry_button.click( fn=delete_prev_fn, inputs=chatbot, outputs=[chatbot, saved_input], api_name=False, queue=False, ).then( fn=display_input, inputs=[saved_input, chatbot], outputs=chatbot, api_name=False, queue=False, ).then( fn=bot, inputs=[ chatbot, max_new_tokens, temperature, top_p, top_k, system_prompt, ], outputs=chatbot, ) undo_button.click( fn=delete_prev_fn, inputs=chatbot, outputs=[chatbot, saved_input], api_name=False, queue=False, ).then( fn=lambda x: x, inputs=[saved_input], outputs=msg, api_name=False, queue=False, ) clear.click(lambda: None, None, chatbot, queue=False) gr.Markdown(LICENSE) demo.queue(max_size=128) demo.launch()