End of training

9017290 verified 8 months ago

2.24 kB

datasets: HectorHe/math7k
library_name: transformers
model_name: Deepseek-V2-13B-Math7K-Expert-Enhance-Subset-Expert-MoE-32-experts
tags:
  - generated_from_trainer
  - open-r1
  - trl
  - expert_enhance_distillation
licence: license

Model Card for Deepseek-V2-13B-Math7K-Expert-Enhance-Subset-Expert-MoE-32-experts

This model is a fine-tuned version of None on the HectorHe/math7k dataset. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="HectorHe/Deepseek-V2-13B-Math7K-Expert-Enhance-Subset-Expert-MoE-32-experts", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with Expert Enhance Trainer, a method introduced in None.

Framework versions

TRL: 0.18.0.dev0
Transformers: 4.52.1
Pytorch: 2.6.0
Datasets: 4.0.0
Tokenizers: 0.21.4

Citations

Cite Expert Enhance Trainer as:

@inproceedings{Github: https://github.com/HectorHHZ/Efficient-Distillation,
    title        = {{Efficient Distillation}},
    author       = {Haoze He(Hector), CMU},
    year         = 2025,
}

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}