You can deploy this model and start earning money today!

RichardErkhov/
bigheiniuJ_-_zephyr-7b-dpo-full-shuffle-bad-after-gguf

Name: bigheiniuJ_-_zephyr-7b-dpo-full-shuffle-bad-after-gguf
Availability: InStock

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

zephyr-7b-dpo-full-shuffle-bad-after - GGUF

Model creator: https://huggingface.co/bigheiniuJ/
Original model: https://huggingface.co/bigheiniuJ/zephyr-7b-dpo-full-shuffle-bad-after/

Name	Quant method	Size
zephyr-7b-dpo-full-shuffle-bad-after.Q2_K.gguf	Q2_K	2.53GB
zephyr-7b-dpo-full-shuffle-bad-after.IQ3_XS.gguf	IQ3_XS	2.81GB
zephyr-7b-dpo-full-shuffle-bad-after.IQ3_S.gguf	IQ3_S	2.96GB
zephyr-7b-dpo-full-shuffle-bad-after.Q3_K_S.gguf	Q3_K_S	2.95GB
zephyr-7b-dpo-full-shuffle-bad-after.IQ3_M.gguf	IQ3_M	3.06GB
zephyr-7b-dpo-full-shuffle-bad-after.Q3_K.gguf	Q3_K	3.28GB
zephyr-7b-dpo-full-shuffle-bad-after.Q3_K_M.gguf	Q3_K_M	3.28GB
zephyr-7b-dpo-full-shuffle-bad-after.Q3_K_L.gguf	Q3_K_L	3.56GB
zephyr-7b-dpo-full-shuffle-bad-after.IQ4_XS.gguf	IQ4_XS	3.67GB
zephyr-7b-dpo-full-shuffle-bad-after.Q4_0.gguf	Q4_0	3.83GB
zephyr-7b-dpo-full-shuffle-bad-after.IQ4_NL.gguf	IQ4_NL	3.87GB
zephyr-7b-dpo-full-shuffle-bad-after.Q4_K_S.gguf	Q4_K_S	3.86GB
zephyr-7b-dpo-full-shuffle-bad-after.Q4_K.gguf	Q4_K	4.07GB
zephyr-7b-dpo-full-shuffle-bad-after.Q4_K_M.gguf	Q4_K_M	4.07GB
zephyr-7b-dpo-full-shuffle-bad-after.Q4_1.gguf	Q4_1	4.24GB
zephyr-7b-dpo-full-shuffle-bad-after.Q5_0.gguf	Q5_0	4.65GB
zephyr-7b-dpo-full-shuffle-bad-after.Q5_K_S.gguf	Q5_K_S	4.65GB
zephyr-7b-dpo-full-shuffle-bad-after.Q5_K.gguf	Q5_K	4.78GB
zephyr-7b-dpo-full-shuffle-bad-after.Q5_K_M.gguf	Q5_K_M	4.78GB
zephyr-7b-dpo-full-shuffle-bad-after.Q5_1.gguf	Q5_1	5.07GB
zephyr-7b-dpo-full-shuffle-bad-after.Q6_K.gguf	Q6_K	5.53GB
zephyr-7b-dpo-full-shuffle-bad-after.Q8_0.gguf	Q8_0	7.17GB

Original model description:

base_model: alignment-handbook/zephyr-7b-sft-full library_name: transformers model_name: zephyr-7b-dpo-full-shuffle-bad-after tags:

generated_from_trainer
trl
dpo licence: license

Model Card for zephyr-7b-dpo-full-shuffle-bad-after

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full. It has been trained using TRL.

Quick start

python

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="bigheiniuJ/zephyr-7b-dpo-full-shuffle-bad-after", 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 DPO, a method introduced in Direct Preference Optimization: Your Language Model is Secretly a Reward Model.

Framework versions

TRL: 0.12.1
Transformers: 4.46.3
Pytorch: 2.5.1
Datasets: 2.21.0
Tokenizers: 0.20.3

Citations

Cite DPO as:

bibtex

@inproceedings{rafailov2023direct,
    title        = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
    author       = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
    year         = 2023,
    booktitle    = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
    url          = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
    editor       = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
}

Cite TRL as: