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

A simple but complete full-attention transformer with a set of promising experimental features from various papers

Re-implement CycleGAN in Tensorlayer

Model-based Reinforcement Learning Improves Autonomous Racing Performance

Code for our paper "Sematic Representation for Dialogue Modeling" in ACL2021

E2e music remastering system - End-to-end Music Remastering System Using Self-supervised and Adversarial Training

Code to reproduce the experiments in the paper "Transformer Based Multi-Source Domain Adaptation" (EMNLP 2020)

Code for "On the Effects of Batch and Weight Normalization in Generative Adversarial Networks"

Simple, efficient and flexible vision toolbox for mxnet framework.

A working implementation of the Categorical DQN (Distributional RL).

Effect of Different Encodings and Distance Functions on Quantum Instance-based Classifiers

Code for: Gradient-based Hierarchical Clustering using Continuous Representations of Trees in Hyperbolic Space. Nicholas Monath, Manzil Zaheer, Daniel Silva, Andrew McCallum, Amr Ahmed. KDD 2019.

Synthetic structured data generators

[SIGIR22] Official PyTorch implementation for "CORE: Simple and Effective Session-based Recommendation within Consistent Representation Space".

Cross View SLAM

Bayesian Neural Networks in PyTorch

Bootstrapped Representation Learning on Graphs

Deploy a ML inference service on a budget in less than 10 lines of code.

Train emoji embeddings based on emoji descriptions.

code and models for "Laplacian Pyramid Reconstruction and Refinement for Semantic Segmentation"

Code for our paper "SimCLS: A Simple Framework for Contrastive Learning of Abstractive Summarization", ACL 2021