Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/THUDM/P-tuning-v2
An optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks
https://github.com/THUDM/P-tuning-v2
natural-language-processing p-tuning parameter-efficient-learning pretrained-language-model prompt-tuning
Last synced: about 2 months ago
JSON representation
An optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks
- Host: GitHub
- URL: https://github.com/THUDM/P-tuning-v2
- Owner: THUDM
- License: apache-2.0
- Created: 2021-10-14T14:16:05.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2023-11-16T04:38:09.000Z (about 1 year ago)
- Last Synced: 2024-10-21T17:19:04.598Z (about 2 months ago)
- Topics: natural-language-processing, p-tuning, parameter-efficient-learning, pretrained-language-model, prompt-tuning
- Language: Python
- Homepage:
- Size: 1.41 MB
- Stars: 1,974
- Watchers: 29
- Forks: 201
- Open Issues: 34
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- Awesome-LLM-Productization - P-tuning v2 - An optimized prompt tuning strategy achieving comparable performance to fine-tuning on small/medium-sized models and sequence tagging challenges; (Models and Tools / LLM Finetuning)
- StarryDivineSky - THUDM/P-tuning-v2 - tuning v2 对预训练变压器的每一层输入应用连续提示。深度提示调整增加了连续提示的容量，并缩小了跨各种设置微调的差距，特别是对于小型模型和艰巨的任务。将文本生成的prefix-tuning技术适配到NLU任务。Prompting技术火爆NLP社区，其将预训练模型从Fine-tuning范式带入Prompt-Engineering时代。Promp最初由人工设计，自然语言提示本身十分脆弱，而且从优化角度无法达到最优。为了解决问题发展出了可学习的Prompt，而P-tuning v2在实际上就是Prefix-tuning，在Prefix部分，每一层transformer的embedding输入需要被tuned。在不同规模大小的LM模型上，P-tuning v2能与精调（Fine-tuning）方法的表现比肩，有时甚至更好。 (预训练模型)
- awesome-llmops - p-tuning-v2 - tuning on small/medium-sized models and sequence tagging challenges. [(ACL 2022)](https://arxiv.org/abs/2110.07602) |  | (Training / Foundation Model Fine Tuning)
README
# P-tuning v2
Source codes and data for
* [ACL 2022] [P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks](https://arxiv.org/abs/2110.07602)
* [Findings of EMNLP 2023] [Parameter-Efficient Prompt Tuning Makes Generalized and Calibrated Neural Text Retrievers](https://arxiv.org/pdf/2207.07087.pdf) [[Code]](https://github.com/THUDM/P-tuning-v2/tree/main/PT-Retrieval)
An optimized prompt tuning strategy achieving comparable performance to fine-tuning on small/medium-sized models and sequence tagging challenges.
Find our previous version [P-tuning v1](https://github.com/THUDM/P-tuning) for knowledge probing and few-shot SuperGLUE. Your kindly starring our repo can greatly encourage us to work harder :)
You may be also interested in our recent work [GLM-130B: An Open Bilingual Pre-trained Model (2022-10-06)](https://arxiv.org/abs/2210.02414). It is an open-sourced LLM outperforming GPT-3 175B over various benchmarks. Get model weights, do inference and P-Tuning v2 with only **4 * RTX 3090 or 8 * RTX 2080 Ti** [FOR FREE](https://github.com/THUDM/GLM-130B)!
P-tuning v2 leverages **deep prompt tuning**, which is to apply continuous prompts for every layer input of the pretrained transformer.
Deep prompt tuning increases the capacity of continuous prompts and closes the gap to fine-tuning across various settings, especially for small models and hard tasks.
Thanks [@rainatam](https://github.com/rainatam)'s joint effort in re-organizing codes for publishing!
## Commonly Asked Question
1. Some readers notice a **'mismatch'** in SuperGLUE between P-tuning (v1) and P-tuning v2: This is because in P-tuning's SuperGLUE experiment, for fair comparison to PET, we follow its experimental setting where backbone pre-trained model parameters are jointly tuned with continuous prompt embeddings; while in P-tuning v2, we follow Prefix tuning and Lester et al.'s parameter-efficient setting where backbone pre-trained model parameters are frozen.
## Reproduce Tips
Since experiments reported in our paper are all conducted on NVIDIA DGX-A100 servers (which might be difficult to acquire),
we reimplement P-tuning v2's results on BERT-large/RoBERTa-large with:
* Ubuntu servers with NVIDIA GeForce RTX 3090 (24G) GPUs
* cuda 11.1
* packages with certain versions (provided below)
We notice that the best hyper-parameters can be sensitive to your server environment and package version.
If you do not have the exact same environment, we highly recommend you to run hyper-parameter search in your environment
based on our example hyper-parameter search script in [search_script](search_script) and result collection scripts [search.py](search.py).
### Setup
We conduct our experiment with Anaconda3. If you have installed Anaconda3, then create the environment for P-tuning v2:
```shell
conda create -n pt2 python=3.8.5
conda activate pt2
```
After we setup basic conda environment, install pytorch related packages via:
```shell
conda install -n pt2 pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=11.0 -c pytorch
```
Finally, install other python packages we need:
```shell
pip install -r requirements.txt
```
### Data
For SuperGLUE and SQuAD datasets, we download them from the Huggingface Datasets APIs (embedded in our codes).
For sequence tagging (NER, SRL) datasets, we prepare a non-official packup [here](https://zenodo.org/record/6318701/files/P-tuning-v2_data.tar.gz?download=1).
After downloading, unzip the packup to the project root.
Please use at your own risk.
### Training
Run training scripts in [run_script](run_script) (e.g., RoBERTa for RTE):
```shell
bash run_script/run_rte_roberta.sh
```
### Implemented Results
Currently we have released our reimplementation on following tasks and datasets. More implementation will be released soon.
Released results on BERT-large
| | BoolQ | COPA | RTE | WiC | WSC | CoNLL04 | OntoNotes 5.0 | CoNLL12 |
|--------------|-------|------|------|------|------|---------|---------------|---------|
| Result | 74.3 | 77.0 | 80.1 | 75.1 | 68.3 | 84.5 | 86.4 | 85.3 |
| Total Epochs | 100 | 80 | 60 | 80 | 80 | 40 | 30 | 45 |
| Best Epoch | 58 | 12 | 30 | 56 | 17 | 33 | 24 | 43 |
Released results on RoBERTa-large
| | BoolQ | COPA | RTE | WiC | WSC | CoNLL03 | CoNLL04 | OntoNotes 5.0 | CoNLL12 | CoNLL05 WSJ | CoNLL05 Brown | SQuAD 1.1 | SQuAD 2.0 |
|--------------|-------|------|------|------|------|---------|---------|---------------|---------|-------------|---------------|-----------|-----------|
| Results | 84.0 | 92.0 | 86.6 | 73.7 | 64.4 | 91.8 | 88.4 | 90.1 | 84.7 | 89.4 | 83.9 | 88.1/94.2 | 81.3/84.7 |
| Total Epochs | 100 | 120 | 100 | 50 | 10 | 30 | 80 | 60 | 45 | 15 | - | 30 | 10 |
| Best Epoch | 86 | 78 | 65 | 31 | 3 | 28 | 45 | 59 | 37 | 13 | - | 24 | 9 |
For other hyper-parameters, please refer to the training scripts.
If you can not achieve the reported results at the best epoch, there is probably an environmental mismatch and hyper-parameter search is needed.
## Citation
If you find our work useful, please kindly cite our paper:
```
@article{DBLP:journals/corr/abs-2110-07602,
author = {Xiao Liu and
Kaixuan Ji and
Yicheng Fu and
Zhengxiao Du and
Zhilin Yang and
Jie Tang},
title = {P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally
Across Scales and Tasks},
journal = {CoRR},
volume = {abs/2110.07602},
year = {2021},
url = {https://arxiv.org/abs/2110.07602},
eprinttype = {arXiv},
eprint = {2110.07602},
timestamp = {Fri, 22 Oct 2021 13:33:09 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2110-07602.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```