Quantization made by Richard Erkhov.

Bookworm-10.7B-v0.4-DPO - GGUF

Model creator: https://huggingface.co/yanolja/
Original model: https://huggingface.co/yanolja/Bookworm-10.7B-v0.4-DPO/

Name	Quant method	Size
Bookworm-10.7B-v0.4-DPO.Q2_K.gguf	Q2_K	3.77GB
Bookworm-10.7B-v0.4-DPO.IQ3_XS.gguf	IQ3_XS	4.18GB
Bookworm-10.7B-v0.4-DPO.IQ3_S.gguf	IQ3_S	4.41GB
Bookworm-10.7B-v0.4-DPO.Q3_K_S.gguf	Q3_K_S	4.39GB
Bookworm-10.7B-v0.4-DPO.IQ3_M.gguf	IQ3_M	4.56GB
Bookworm-10.7B-v0.4-DPO.Q3_K.gguf	Q3_K	4.88GB
Bookworm-10.7B-v0.4-DPO.Q3_K_M.gguf	Q3_K_M	4.88GB
Bookworm-10.7B-v0.4-DPO.Q3_K_L.gguf	Q3_K_L	5.31GB
Bookworm-10.7B-v0.4-DPO.IQ4_XS.gguf	IQ4_XS	5.47GB
Bookworm-10.7B-v0.4-DPO.Q4_0.gguf	Q4_0	5.7GB
Bookworm-10.7B-v0.4-DPO.IQ4_NL.gguf	IQ4_NL	5.77GB
Bookworm-10.7B-v0.4-DPO.Q4_K_S.gguf	Q4_K_S	5.75GB
Bookworm-10.7B-v0.4-DPO.Q4_K.gguf	Q4_K	6.07GB
Bookworm-10.7B-v0.4-DPO.Q4_K_M.gguf	Q4_K_M	6.07GB
Bookworm-10.7B-v0.4-DPO.Q4_1.gguf	Q4_1	6.32GB
Bookworm-10.7B-v0.4-DPO.Q5_0.gguf	Q5_0	6.94GB
Bookworm-10.7B-v0.4-DPO.Q5_K_S.gguf	Q5_K_S	6.94GB
Bookworm-10.7B-v0.4-DPO.Q5_K.gguf	Q5_K	7.13GB
Bookworm-10.7B-v0.4-DPO.Q5_K_M.gguf	Q5_K_M	7.13GB
Bookworm-10.7B-v0.4-DPO.Q5_1.gguf	Q5_1	7.56GB
Bookworm-10.7B-v0.4-DPO.Q6_K.gguf	Q6_K	8.26GB
Bookworm-10.7B-v0.4-DPO.Q8_0.gguf	Q8_0	10.69GB

Original model description:

license: apache-2.0 base_model: yanolja/KoSOLAR-10.7B-v0.2 tags: - generated_from_trainer model-index: - name: yanolja/Bookworm-10.7B-v0.4-DPO results: []

Bookworm-10.7B-v0.4-DPO

Join Our Community on Discord!

If you're passionate about the field of Large Language Models and wish to exchange knowledge and insights, we warmly invite you to join our Discord server. It's worth noting that Korean is the primary language used in this server. The landscape of LLM is evolving rapidly, and without active sharing, our collective knowledge risks becoming outdated swiftly. Let's collaborate and drive greater impact together! Join us here: Discord Link.

Our Dedicated Team (Alphabetical Order)

Research	Engineering	Product Management	UX Design
Myeongho Jeong	Geon Kim	Bokyung Huh	Eunsue Choi
Seungduk Kim	Rifqi Alfi
Seungtaek Choi	Sanghoon Han
	Suhyun Kang

About the Model

This model is a fine-tuned version of yanolja/KoSOLAR-10.7B-v0.2, which is a Korean vocabulary-extended version of upstage/SOLAR-10.7B-v1.0. Specifically, we employed Direct Preference Optimization (DPO) based on LLaMA-Factory.

Training Data

Korean-translated version of Open-Orca/SlimOrca-Dedup
Korean-translated version of argilla/ultrafeedback-binarized-preferences-cleaned
No other dataset was used

Citation

@misc{cui2023ultrafeedback,
      title={UltraFeedback: Boosting Language Models with High-quality Feedback}, 
      author={Ganqu Cui and Lifan Yuan and Ning Ding and Guanming Yao and Wei Zhu and Yuan Ni and Guotong Xie and Zhiyuan Liu and Maosong Sun},
      year={2023},
      eprint={2310.01377},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{SlimOrcaDedup,
  title = {SlimOrca Dedup: A Deduplicated Subset of SlimOrca},
  author = {Wing Lian and Guan Wang and Bleys Goodson and Eugene Pentland and Austin Cook and Chanvichet Vong and "Teknium" and Nathan Hoos},
  year = {2023},
  publisher = {HuggingFace},
  url = {https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup/}
}

@misc{mukherjee2023orca,
      title={Orca: Progressive Learning from Complex Explanation Traces of GPT-4}, 
      author={Subhabrata Mukherjee and Arindam Mitra and Ganesh Jawahar and Sahaj Agarwal and Hamid Palangi and Ahmed Awadallah},
      year={2023},
      eprint={2306.02707},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}