Implementation of ICLR 2020 paper "Revisiting Self-Training for Neural Sequence Generation"

Overview

Self-Training for Neural Sequence Generation

This repo includes instructions for running noisy self-training algorithms from the following paper:

Revisiting Self-Training for Neural Sequence Generation
Junxian He*, Jiatao Gu*, Jiajun Shen, Marc'Aurelio Ranzato
ICLR 2020

Requirement

  • fairseq (please see the fairseq repo for other requirements on Python and PyTorch versions)

fairseq can be installed with:

pip install fairseq

Data

Download and preprocess the WMT'14 En-De dataset:

# Download and prepare the data
wget https://raw.githubusercontent.com/pytorch/fairseq/master/examples/translation/prepare-wmt14en2de.sh
bash prepare-wmt14en2de.sh --icml17

TEXT=wmt14_en_de
fairseq-preprocess --source-lang en --target-lang de \
    --trainpref $TEXT/train --validpref $TEXT/valid --testpref $TEXT/test \
    --destdir wmt14_en_de_bin --thresholdtgt 0 --thresholdsrc 0 \
    --joined-dictionary --workers 16

Then we mimic a semi-supervised setting where 100K training samples are randomly selected as parallel corpus and the remaining English training samples are treated as unannotated monolingual corpus:

bash extract_wmt100k.sh

Preprocess WMT100K:

bash preprocess.sh 100ken 100kde 

Add noise to the monolingual corpus for later usage:

TEXT=wmt14_en_de
python paraphrase/paraphrase.py \
  --paraphraze-fn noise_bpe \
  --word-dropout 0.2 \
  --word-blank 0.2 \
  --word-shuffle 3 \
  --data-file ${TEXT}/train.mono_en \
  --output ${TEXT}/train.mono_en_noise \
  --bpe-type subword

Train the base supervised model

Train the translation model with 30K updates:

bash supervised_train.sh 100ken 100kde 30000

Self-training as pseudo-training + fine-tuning

Translate the monolingual data to train.[suffix] to form a pseudo parallel dataset:

bash translate.sh [model_path] [suffix]  

Suppose the pseduo target language suffix is mono_de_iter1 (by default), preprocess:

bash preprocess.sh mono_en_noise mono_de_iter1

Pseudo-training + fine-tuning:

bash self_train.sh mono_en_noise mono_de_iter1 

The above command trains the model on the pseduo parallel corpus formed by train.mono_en_noise and train.mono_de_iter1 and then fine-tune it on real parallel data.

This self-training process can be repeated for multiple iterations to improve performance.

Reference

@inproceedings{He2020Revisiting,
title={Revisiting Self-Training for Neural Sequence Generation},
author={Junxian He and Jiatao Gu and Jiajun Shen and Marc'Aurelio Ranzato},
booktitle={Proceedings of ICLR},
year={2020},
url={https://openreview.net/forum?id=SJgdnAVKDH}
}
You might also like...
UDP++ (ECCVW 2020 Oral), (Winner of COCO 2020 Keypoint Challenge).
UDP++ (ECCVW 2020 Oral), (Winner of COCO 2020 Keypoint Challenge).

UDP-Pose This is the pytorch implementation for UDP++, which won the Fisrt place in COCO Keypoint Challenge at ECCV 2020 Workshop. Top-Down Results on

PyTorch code for ICLR 2021 paper Unbiased Teacher for Semi-Supervised Object Detection
PyTorch code for ICLR 2021 paper Unbiased Teacher for Semi-Supervised Object Detection

Unbiased Teacher for Semi-Supervised Object Detection This is the PyTorch implementation of our paper: Unbiased Teacher for Semi-Supervised Object Detection

Seach Losses of our paper 'Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search', accepted by ICLR 2021.
Seach Losses of our paper 'Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search', accepted by ICLR 2021.

CSE-Autoloss Designing proper loss functions for vision tasks has been a long-standing research direction to advance the capability of existing models

Code for ICLR 2021 Paper,
Code for ICLR 2021 Paper, "Anytime Sampling for Autoregressive Models via Ordered Autoencoding"

Anytime Autoregressive Model Anytime Sampling for Autoregressive Models via Ordered Autoencoding , ICLR 21 Yilun Xu, Yang Song, Sahaj Gara, Linyuan Go

Code for the paper
Code for the paper "Training GANs with Stronger Augmentations via Contrastive Discriminator" (ICLR 2021)

Training GANs with Stronger Augmentations via Contrastive Discriminator (ICLR 2021) This repository contains the code for reproducing the paper: Train

Official code for the ICLR 2021 paper Neural ODE Processes
Official code for the ICLR 2021 paper Neural ODE Processes

Neural ODE Processes Official code for the paper Neural ODE Processes (ICLR 2021). Abstract Neural Ordinary Differential Equations (NODEs) use a neura

Source code, datasets and trained models for the paper Learning Advanced Mathematical Computations from Examples (ICLR 2021), by François Charton, Amaury Hayat (ENPC-Rutgers) and Guillaume Lample

Maths from examples - Learning advanced mathematical computations from examples This is the source code and data sets relevant to the paper Learning a

Based on the paper
Based on the paper "Geometry-aware Instance-reweighted Adversarial Training" ICLR 2021 oral

Geometry-aware Instance-reweighted Adversarial Training This repository provides codes for Geometry-aware Instance-reweighted Adversarial Training (ht

Official repository for the ICLR 2021 paper Evaluating the Disentanglement of Deep Generative Models with Manifold Topology

Official repository for the ICLR 2021 paper Evaluating the Disentanglement of Deep Generative Models with Manifold Topology Sharon Zhou, Eric Zelikman

Comments
  • supervised_training error

    supervised_training error

    Hi Junxian!

    I encountered this error when I tried to run supervised_train.sh OSError: Model file not found: checkpoint/transformer_wmt_en_de.drop0.3.100ken-100kde.nsteps30000/checkpoint_best.pt Any ideas? Thank you.

    opened by haorannlp 2
Owner
Junxian He
NLP/ML PhD student at CMU
Junxian He
Pytorch code for "State-only Imitation with Transition Dynamics Mismatch" (ICLR 2020)

This repo contains code for our paper State-only Imitation with Transition Dynamics Mismatch published at ICLR 2020. The code heavily uses the RL mach

null 20 Sep 8, 2022
A supplementary code for Editable Neural Networks, an ICLR 2020 submission.

Editable neural networks A supplementary code for Editable Neural Networks, an ICLR 2020 submission by Anton Sinitsin, Vsevolod Plokhotnyuk, Dmitry Py

Anton Sinitsin 32 Nov 29, 2022
Official implementation of the ICLR 2021 paper

You Only Need Adversarial Supervision for Semantic Image Synthesis Official PyTorch implementation of the ICLR 2021 paper "You Only Need Adversarial S

Bosch Research 272 Dec 28, 2022
This project is the official implementation of our accepted ICLR 2021 paper BiPointNet: Binary Neural Network for Point Clouds.

BiPointNet: Binary Neural Network for Point Clouds Created by Haotong Qin, Zhongang Cai, Mingyuan Zhang, Yifu Ding, Haiyu Zhao, Shuai Yi, Xianglong Li

Haotong Qin 59 Dec 17, 2022
An implementation demo of the ICLR 2021 paper Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks in PyTorch.

Neural Attention Distillation This is an implementation demo of the ICLR 2021 paper Neural Attention Distillation: Erasing Backdoor Triggers from Deep

Yige-Li 84 Jan 4, 2023
Official Pytorch implementation of ICLR 2018 paper Deep Learning for Physical Processes: Integrating Prior Scientific Knowledge.

Deep Learning for Physical Processes: Integrating Prior Scientific Knowledge: Official Pytorch implementation of ICLR 2018 paper Deep Learning for Phy

emmanuel 47 Nov 6, 2022
Tensorflow 2 implementation of the paper: Learning and Evaluating Representations for Deep One-class Classification published at ICLR 2021

Deep Representation One-class Classification (DROC). This is not an officially supported Google product. Tensorflow 2 implementation of the paper: Lea

Google Research 137 Dec 23, 2022
Pytorch Implementation of Continual Learning With Filter Atom Swapping (ICLR'22 Spolight) Paper

Continual Learning With Filter Atom Swapping Pytorch Implementation of Continual Learning With Filter Atom Swapping (ICLR'22 Spolight) Paper If find t

null 11 Aug 29, 2022
ReLoss - Official implementation for paper "Relational Surrogate Loss Learning" ICLR 2022

Relational Surrogate Loss Learning (ReLoss) Official implementation for paper "R

Tao Huang 31 Nov 22, 2022
A PyTorch implementation of ICLR 2022 Oral paper PiCO: Contrastive Label Disambiguation for Partial Label Learning

PiCO: Contrastive Label Disambiguation for Partial Label Learning This is a PyTorch implementation of ICLR 2022 Oral paper PiCO; also see our Project

王皓波 83 May 11, 2022