Prefix Tuning

Files:

.
├── gpt2                          # Code for GPT2 style autoregressive LM
│   ├── train_e2e.py              # high-level scripts to train.
│   ├── train_control.py          # code that implements prefix-tuning.
│   ├── trainer_prefix.py         # trainer code for the training loop. 
│   ├── run_language_modeling.py  # training code (contains data loading, model loading, and calls trainer)
│   ├── gen.py                    # high-level scripts to decode. 
│   └── run_generation.py         # decoding code. 
│
├── seq2seq                       # Code for encoder-decoder architecture
│   ├── train_bart.py             # high-level scripts to train.
│   ├── prefixTuning.py           # code that implements prefix-tuning.
│   ├── finetune.py               # training code (contains data loading, model loading, and calls trainer)   
│   ├── lightning_base.py         # helper code
│   ├── utils.py                  # helper code
│   └── callbacks.py              # helper code
└── ...

To run the code for GPT2 style autoregressive LM, the code is in gpt2/. This corresponds to the table-to-text experiments in the paper.

To run the code for encoder-decoder architecture like BART, the code is in seq2seq. This corresponds to the summarization experiments in the paper.

The two primary scripts I used to run my codes are gpt2/train_e2e.py (for table-to-text) and seq2seq/train_bart.py(for summarization). they are set to default of good hyperparameters, and can be used to tune hyperparameter :)

Setup:

cd transformer; pip install -e .

Train via prefix-tuning:

cd gpt2;

python train_e2e.py --optim_prefix yes --preseqlen 5 --epoch 5 --learning_rate 0.00005 --mode webnlg --bsz 5 --seed 101

cd seq2seq; 

python train_bart.py --mode xsum --preseqlen 200 --do_train yes --fp16 yes --bsz 16  --epoch 30  --gradient_accumulation_step 3 --learning_rate 0.00005  --mid_dim 800

Other baseline approaches

cd gpt2;

python train_e2e.py --tuning_mode {finetune/adaptertune} --epoch 5 --learning_rate 0.00005 --mode webnlg --bsz 5 --seed 101

cd seq2seq;

python train_e2e.py --tuning_mode finetune --epoch 5 --learning_rate 0.00005 --mode webnlg --bsz 5 --seed 101

Decode:

cd gpt2;

python gen.py {data2text/webnlg/...} yes test {checkpoint_path} no

cd seq2seq; 

python train_bart.py --mode xsum --do_train no --prefix_model_path {checkpoint_path} --preseqlen {same as training} --mid_dim {same as training}

For details of the methods and results, please refer to our paper.

@misc{li2021prefixtuning,
      title={Prefix-Tuning: Optimizing Continuous Prompts for Generation}, 
      author={Xiang Lisa Li and Percy Liang},
      year={2021},
      eprint={2101.00190},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Prefix-Tuning: Optimizing Continuous Prompts for Generation

Related tags

Overview

Prefix Tuning

Files:

Setup:

Train via prefix-tuning:

Decode:

Owner

Highly comparative time-series analysis

This repository contains numerical implementation for the paper Intertemporal Pricing under Reference Effects: Integrating Reference Effects and Consumer Heterogeneity.

GemNet model in PyTorch, as proposed in "GemNet: Universal Directional Graph Neural Networks for Molecules" (NeurIPS 2021)

Python Wrapper for Embree

Implementation of BI-RADS-BERT & The Advantages of Section Tokenization.

This is the code repository for the paper A hierarchical semantic segmentation framework for computer-vision-based bridge column damage detection

banditml is a lightweight contextual bandit & reinforcement learning library designed to be used in production Python services.

Official repository for MixFaceNets: Extremely Efficient Face Recognition Networks

Official implementation for TTT++: When Does Self-supervised Test-time Training Fail or Thrive

Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes, ICCV 2017

An efficient framework for reinforcement learning.

Combine Tacotron2 and Hifi GAN to generate speech from text

An end-to-end implementation of intent prediction with Metaflow and other cool tools

Relaxed-machines - explorations in neuro-symbolic differentiable interpreters

Emotion classification of online comments based on RNN

Code for "Learning Graph Cellular Automata"

Code for ACL 2019 Paper: "COMET: Commonsense Transformers for Automatic Knowledge Graph Construction"

Official implementation of the paper ``Unifying Nonlocal Blocks for Neural Networks'' (ICCV'21)

Discovering Explanatory Sentences in Legal Case Decisions Using Pre-trained Language Models.

A project for developing transformer-based models for clinical relation extraction