app.py

import spaces
import re 
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
import torch

title = """# 🙋🏻‍♂️Welcome to🌟Tonic's🔮DeepSeekMath📉
You can build with this endpoint using🔮DeepSeekMath📉 available here : [deepseek-ai/deepseek-math-7b-instruct](https://huggingface.co/deepseek-ai/deepseek-math-7b-instruct). We're using 🤖[introspector/unimath](https://huggingface.co/datasets/introspector/unimath) for cool examples, check it out below ! The demo is still a work in progress and we're looking forward to build downstream tasks that showcase outstanding mathematical reasoning. Have any ideas ? join us below !
You can also use 🔮DeepSeekMath📉 by cloning this space. Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic/Math?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: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) Math with [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 🤗
"""

additional_info_prompt = "Explain the above using mathematics, print entire answer in latex format:"

unimath1 = """Goal:
  
  X : UU
  Y : UU
  P : UU
  xp : (X → P) → P
  yp : (Y → P) → P
  X0 : X × Y → P
  x : X
  ============================
   (Y → P)


DEBUG:Going to execute:
PTRDEBUGTAC<coq-core.plugins.ltac::intro@1> $1
DEBUG LTAC Evaluated term: yp

TcDebug (0) > NONE??? LtacNotationCall (no location),/mnt/data1/2024/01/05/UniMath/UniMath/Foundations/PartA.v:234
"""
# source : unimath/unimath/batch2/data08

unimath2 = """Goal:
    R : ring  M : module R
  ============================
   (islinear (idfun M))


DEBUG:Going to execute:
PTRDEBUGTACapply pathsinv0; trivial
Level 0: Backtrace:

Proof is not complete.
Level 0: Backtrace:

Proof is not complete.

TcDebug (0) > NONE??? LtacNotationCall (no location),/mnt/data1/2024/01/05/UniMath/UniMath/MoreFoundations/Tactics.veasy:19, Tactics  (UniMath.MoreFoundations),/mnt/data1/2024/01/05/UniMath/UniMath/Algebra/Modules/Examples.v:27

"""
# source : unimath/unimath/batch2/data_22/BATCH122007

unimath3 = """Goal:
    X : UU  i : nat  b : hProptoType (i < S i)  x : Vector X (S i)  r : i = i
  ============================
   (pr1 lastelement = pr1 (i,, b))


DEBUG:Going to execute:
PTRDEBUGTACsimpl
DEBUG LTAC Evaluated term: isinjstntonat

TcDebug (0) > /mnt/data1/2024/01/05/UniMath/UniMath/Combinatorics/FiniteSequences.v:114
"""
# source : unimath/unimath/batch2/data_12/BATCH112026

unimath4 = """Goal:
    X : dcpo  CX : continuous_dcpo_struct X  x : pr1hSet X  y : pr1hSet X
  ============================
   (x ⊑ y
    ≃ (∀ i : approximating_family CX x, approximating_family CX x i ⊑ y))


DEBUG:Going to execute:
PTRDEBUGTACsimple refine (p _ _ _) ||
             simple refine (p _ _ _ _) ||
               simple refine (p _ _ _ _ _) ||
                 simple refine (p _ _ _ _ _ _) ||
                   simple refine (p _ _ _ _ _ _ _) ||
                     simple refine (p _ _ _ _ _ _ _ _) ||
                       simple refine (p _ _ _ _ _ _ _ _ _) ||
                         simple refine (p _ _ _ _ _ _ _ _ _ _) ||
                           simple refine (p _ _ _ _ _ _ _ _ _ _ _) ||
                             simple refine (p _ _ _ _ _ _ _ _ _ _ _ _) ||
                               simple refine (p _ _ _ _ _ _ _ _ _ _ _ _ _) ||
                                 simple refine
                                  (p _ _ _ _ _ _ _ _ _ _ _ _ _ _) || simple
                                  refine (p _ _ _ _ _ _ _ _ _ _ _ _ _ _ _)
Level 0: Backtrace:

In environment
X : dcpo
CX : continuous_dcpo_struct X
x, y : X
The term "weqimplimpl ?f ?g" has type "isaprop ?X → isaprop ?Y → ?X ≃ ?Y"
while it is expected to have type
 "x ⊑ y ≃ (∀ i : approximating_family CX x, approximating_family CX x i ⊑ y)".
Level 0: Backtrace:

In environment
X : dcpo
CX : continuous_dcpo_struct X
x, y : X
The term "weqimplimpl ?f ?g" has type "isaprop ?X → isaprop ?Y → ?X ≃ ?Y"
while it is expected to have type
 "x ⊑ y ≃ (∀ i : approximating_family CX x, approximating_family CX x i ⊑ y)".

TcDebug (0) > NONE??? LtacNotationCall (no location),/mnt/data1/2024/01/05/UniMath/UniMath/Foundations/Init.vsimple_rapply:174, Init  (UniMath.Foundations),/mnt/data1/2024/01/05/UniMath/UniMath/Foundations/Init.v??? LtacNotationCall:189, Init  (UniMath.Foundations),/mnt/data1/2024/01/05/UniMath/UniMath/OrderTheory/DCPOs/Basis/Continuous.v:166
"""
# source : unimath/unimath/batch2/data_42/BATCH142042

examples = [
    [unimath1, additional_info_prompt, 1200],
    [unimath2, additional_info_prompt, 1200],
    [unimath3, additional_info_prompt, 1200],
    [unimath4, additional_info_prompt, 1200]
]

model_name = "deepseek-ai/deepseek-math-7b-instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
model.generation_config = GenerationConfig.from_pretrained(model_name)
model.generation_config.pad_token_id = model.generation_config.eos_token_id

def parse_full_answer(answer):
    """Parses the assistant's answer, excluding any text before 'Assistant :'."""
    match = re.search(r"Assistant\s*:\s*(.*)", answer, re.DOTALL)
    return match.group(1).strip() if match else "No assistant answer found."

def parse_final_answer(answer):
    """Extracts the final answer enclosed within \boxed{}."""
    match = re.search(r"\\boxed\{([^}]+)\}", answer)
    return match.group(1) if match else "No final answer found."

@spaces.GPU
def solve_math_problem(question, additional_info, max_tokens):
    prompt = f"User: {question}\n{additional_info}.\nAssistant:"
    input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(model.device)
    outputs = model.generate(input_ids, max_length=max_tokens + input_ids.shape[1], pad_token_id=model.generation_config.pad_token_id)
    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
    full_answer = parse_full_answer(result)
    final_answer = parse_final_answer(result)
    return full_answer, final_answer


def main():
    iface = gr.Interface(        
        title="👋🏻Welcome to🌟Tonic's 🔮DeepSeekMath📉",
        description="""You can build with this endpoint using🔮DeepSeekMath📉 available here : [deepseek-ai/deepseek-math-7b-instruct](https://huggingface.co/deepseek-ai/deepseek-math-7b-instruct). We're using 🤖[introspector/unimath](https://huggingface.co/datasets/introspector/unimath) for cool examples, check it out below ! The demo is still a work in progress and we're looking forward to build downstream tasks that showcase outstanding mathematical reasoning. Have any ideas ? join us below !
You can also use 🔮DeepSeekMath📉 by cloning this space. Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic/Math?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: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) Math with [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 🤗""",
        fn=solve_math_problem,
        outputs=[
            gr.Code(label="🔮TonicsMathAssistant📉", interactive=False),
            gr.Textbox(label="Final Answer")
        ],
        inputs=[
            gr.Textbox(label="🤔Enter your math problem", lines=7),
            gr.Textbox(default=additional_info_prompt, label="🪜Optional train-of-thought"),
            gr.Slider(minimum=150, maximum=1200, default=250, label="🪙Max Tokens")
        ],
        examples=examples
    )

    iface.launch()

if __name__ == "__main__":
    main()