app.py

import spaces
import gradio as gr
import torch
import transformers
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import accelerate
import os

title = """# Welcome to 🌟Tonic's🐇🥷🏻Trinity
You can build with this endpoint using🐇🥷🏻Trinity available here : [WhiteRabbitNeo/Trinity-13B](https://huggingface.co.//WhiteRabbitNeo/Trinity-13B). You can also use 🐇🥷🏻Trinity by cloning this space. Simply click here: <a style="display:inline-block" href="https://huggingface.co./spaces/Tonic/trinity?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a></h3> 
Join us : 🌟TeamTonic🌟 is always making cool demos! Join our active builder's 🛠️community 👻 [![Join us on Discord](https://img.shields.io/discord/1109943800132010065?label=Discord&logo=discord&style=flat-square)](https://discord.gg/GWpVpekp) On 🤗Huggingface:[MultiTransformer](https://huggingface.co./MultiTransformer) Math 🔍 [introspector](https://huggingface.co./introspector) On 🌐Github: [Tonic-AI](https://github.com/tonic-ai) & contribute to🌟 [SciTonic](https://github.com/Tonic-AI/scitonic)🤗Big thanks to Yuvi Sharma and all the folks at huggingface for the community grant 🤗
"""


default_system_prompt = """
Answer the Question by exploring multiple reasoning paths as follows:
- First, carefully analyze the question to extract the key information components and break it down into logical sub-questions. This helps set up the framework for reasoning. The goal is to construct an internal search tree.
- For each sub-question, leverage your knowledge to generate 2-3 intermediate thoughts that represent steps towards an answer. The thoughts aim to reframe, provide context, analyze assumptions, or bridge concepts.
- Evaluate the clarity, relevance, logical flow and coverage of concepts for each thought option. Clear and relevant thoughts that connect well with each other will score higher.
- Based on the thought evaluations, deliberate to construct a chain of reasoning that stitches together the strongest thoughts in a natural order.
- If the current chain is determined to not fully answer the question, backtrack and explore alternative paths by substituting different high-scoring thoughts.
- Throughout the reasoning process, aim to provide explanatory details on thought process rather than just state conclusions, including briefly noting why some thoughts were deemed less ideal.
- Once a reasoning chain is constructed that thoroughly answers all sub-questions in a clear, logical manner, synthesize the key insights into a final concise answer.
- Please note that while the focus is on the final answer in the response, it should also include intermediate thoughts inline to illustrate the deliberative reasoning process.
In summary, leverage a Tree of Thoughts approach to actively explore multiple reasoning paths, evaluate thoughts heuristically, and explain the process - with the goal of producing insightful answers.
"""
quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.bfloat16
)

model_path = "WhiteRabbitNeo/Trinity-13B"

hf_token = os.getenv("HF_TOKEN")
if not hf_token:
    raise ValueError("Hugging Face token not found. Please set the HF_TOKEN environment variable.")

model = AutoModelForCausalLM.from_pretrained(
    model_path,
     device_map="auto",
     trust_remote_code=True,
     quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)

@spaces.GPU
def generate_text(custom_prompt, user_input, temperature, generate_len, top_p, top_k):
    system_prompt = custom_prompt if custom_prompt else default_system_prompt
    llm_prompt = f"{system_prompt} \nUSER: {user_input} \nASSISTANT: "
    
    tokens = tokenizer.encode(llm_prompt, return_tensors="pt")
    tokens = tokens.to("cuda")

    length = tokens.shape[1]
    with torch.no_grad():
        output = model.generate(
            input_ids=tokens,
            max_length=length + generate_len,
            temperature=temperature,
            top_p=top_p,
            top_k=top_k,
            num_return_sequences=1,
        )
    generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
    answer = generated_text[len(llm_prompt):].strip()
    
    return answer

def gradio_app():
    with gr.Blocks() as demo:
        gr.Markdown(title)
        with gr.Row():
            custom_prompt = gr.Textbox(label="Custom System Prompt (optional)", placeholder="Leave blank to use the default prompt...")
            instruction = gr.Textbox(label="Your Instruction", placeholder="Type your question here...")
        with gr.Row():
            temperature = gr.Slider(minimum=0.1, maximum=1.0, step=0.1, value=0.5, label="Temperature")
            generate_len = gr.Slider(minimum=100, maximum=1024, step=10, value=100, label="Generate Length")
            top_p = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, value=1.0, label="Top P")
            top_k = gr.Slider(minimum=0, maximum=100, step=1, value=50, label="Top K")
        with gr.Row():
            generate_btn = gr.Button("Generate")
        output = gr.Textbox(label="Generated Text", lines=10, placeholder="Generated answer will appear here...")

        generate_btn.click(
            fn=generate_text,
            inputs=[custom_prompt, instruction, temperature, generate_len, top_p, top_k],
            outputs=output
        )

    demo.launch()

if __name__ == "__main__":
    gradio_app()