Fast-OpenMath-Nemotron-14B

Fast-OpenMath-Nemotron-14B

This model is based on the paper A Practical Two-Stage Recipe for Mathematical LLMs: Maximizing Accuracy with SFT and Efficiency with Reinforcement Learning.

By applying SFT and GRPO on difficult math problems, we enhanced the performance of DeepSeek-R1-Distill-Qwen-14B and developed Fast-Math-R1-14B, which achieves approx. 30% faster inference on average, while maintaining accuracy.

In addition, we trained and open-sourced Fast-OpenMath-Nemotron-14B, an efficiency-optimized version of NVIDIA’s OpenMath-Nemotron-14B, following the same approach. Compared to OpenMath-Nemotron-14B, this model enables approx. 30% faster inference on average, with minimal loss in performance.

Technical details can be found in our github repository. Project page: https://analokmaus.github.io/Fast-Math-R1/

Note: This model likely inherits the ability to perform inference in TIR mode from the original model. However, all of our experiments were conducted in CoT mode, and its performance in TIR mode has not been evaluated.

Evaluation

<img src='https://github.com/analokmaus/kaggle-aimo2-fast-math-r1/blob/master/assets/pass1_aime_all.png?raw=true' max-height='400px'>

Inference

vLLM

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer


model_path = 'RabotniKuma/Fast-OpenMath-Nemotron-14B'
vllm_engine = LLM(
    model=model_path,
    max_model_len=8192,
    gpu_memory_utilization=0.9,
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_path)


sampling_params = SamplingParams(
    temperature=1.0,
    top_p=0.90,
    min_p=0.05,
    max_tokens=8192,
    stop='</think>',  # For even faster inference, applying early stopping at the </think> tag and extracting the final boxed content is recommended.
)
messages = [
    {
        'role': 'user',
        'content': (
            'Solve the problem, and put the answer in \boxed{{}}. '
            'Sarah is twice as old as her youngest brother. If the difference between their ages is 15 years. How old is her youngest brother?'
        )
    }
]
messages = tokenizer.apply_chat_template(
    conversation=messages,
    tokenize=False,
    add_generation_prompt=True
)
response = vllm_engine.generate(messages, sampling_params=sampling_params)