|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- databricks/databricks-dolly-15k |
|
language: |
|
- en |
|
metrics: |
|
- rouge |
|
base_model: |
|
- facebook/opt-6.7B |
|
pipeline_tag: text-generation |
|
--- |
|
# KD-OPT-6.7B |
|
|
|
[paper](https://arxiv.org/abs/2306.08543) | [code](https://github.com/microsoft/LMOps/tree/main/minillm) |
|
|
|
**KD-OPT-6.7B** is a model distilled from [OPT-13B](https://huggingface.co./MiniLLM/teacher-OPT-13B) on [databricks-dolly-15k](https://huggingface.co./datasets/aisquared/databricks-dolly-15k) with token-level forward KLD. |
|
|
|
It is used as a baseline for [MiniLLM](https://huggingface.co./MiniLLM/MiniLLM-OPT-6.7B). |
|
|
|
|
|
## Other Baselines |
|
+ [SFT w/o KD](https://huggingface.co./MiniLLM/SFT-OPT-6.7B) |
|
+ [SeqKD](https://huggingface.co./MiniLLM/SeqKD-OPT-6.7B) |
|
|
|
|
|
## Citation |
|
``` |
|
@inproceedings{minillm, |
|
title={MiniLLM: Knowledge Distillation of Large Language Models}, |
|
author={Gu, Yuxian and Dong, Li and Wei, Furu and Huang, Minlie}, |
|
booktitle={Proceedings of ICLR}, |
|
year={2024} |
|
} |
|
``` |
