"Structure-Augmented Text Representation Learning for Efficient Knowledge Graph Completion"(WWW 2021)

Related tags

Deep LearningStAR_KGC
Overview

STAR_KGC

This repo contains the source code of the paper accepted by WWW'2021. "Structure-Augmented Text Representation Learning for Efficient Knowledge Graph Completion"(WWW 2021).

1. Thanks

The repository is partially based on huggingface transformers, KG-BERT and RotatE.

2. Installing requirement packages

  • conda create -n StAR python=3.6
  • source activate StAR
  • pip install numpy torch tensorboardX tqdm boto3 requests regex sacremoses sentencepiece matplotlib
2.1 Optional package (for mixed float Computation)

3. Dataset

  • WN18RR, FB15k-237, UMLS

    • Train and test set in ./data
    • As validation on original dev set is costly, we validated the model on dev subset during training.
    • The dev subset of WN18RR is provided in ./data/WN18RR called new_dev.dict. Use below commands to get the dev subset for WN18RR (FB15k-237 is similar without the --do_lower_case) used in training process.
     CUDA_VISIBLE_DEVICES=0 \
  python get_new_dev_dict.py \
 	--model_class bert \
 	--weight_decay 0.01 \
 	--learning_rate 5e-5 \
 	--adam_epsilon 1e-6 \
 	--max_grad_norm 0. \
 	--warmup_proportion 0.05 \
 	--do_train \
 	--num_train_epochs 7 \
 	--dataset WN18RR \
 	--max_seq_length 128 \
 	--gradient_accumulation_steps 4 \
 	--train_batch_size 16 \
 	--eval_batch_size 128 \
 	--logging_steps 100 \
 	--eval_steps -1 \
 	--save_steps 2000 \
 	--model_name_or_path bert-base-uncased \
 	--do_lower_case \
 	--output_dir ./result/WN18RR_get_dev \
 	--num_worker 12 \
 	--seed 42 \

     CUDA_VISIBLE_DEVICES=0 \
  python get_new_dev_dict.py \
 	--model_class bert \
 	--weight_decay 0.01 \
 	--learning_rate 5e-5 \
 	--adam_epsilon 1e-6 \
 	--max_grad_norm 0. \
 	--warmup_proportion 0.05 \
 	--do_eval \
 	--num_train_epochs 7 \
 	--dataset WN18RR \
 	--max_seq_length 128 \
 	--gradient_accumulation_steps 4 \
 	--train_batch_size 16 \
 	--eval_batch_size 128 \
 	--logging_steps 100 \
 	--eval_steps 1000 \
 	--save_steps 2000 \
 	--model_name_or_path ./result/WN18RR_get_dev \
 	--do_lower_case \
 	--output_dir ./result/WN18RR_get_dev \
 	--num_worker 12 \
 	--seed 42 \

  • NELL-One

    • We reformat original NELL-One as the three benchmarks above.
    • Please run the below command to get the reformatted data.
     python reformat_nell_one.py --data_dir path_to_downloaded --output_dir ./data/NELL_standard

4. Training and Test (StAR)

Run the below commands for reproducing results in paper. Note, all the eval_steps is set to -1 to train w/o validation and save the last checkpoint, because standard dev is very time-consuming. This can get similar results as in the paper.

4.1 WN18RR

CUDA_VISIBLE_DEVICES=0 \
python run_link_prediction.py \
    --model_class roberta \
    --weight_decay 0.01 \
    --learning_rate 1e-5 \
    --adam_betas 0.9,0.98 \
    --adam_epsilon 1e-6 \
    --max_grad_norm 0. \
    --warmup_proportion 0.05 \
    --do_train --do_eval \
    --do_prediction \
    --num_train_epochs 7 \
    --dataset WN18RR \
    --max_seq_length 128 \
    --gradient_accumulation_steps 4 \
    --train_batch_size 16 \
    --eval_batch_size 128 \
    --logging_steps 100 \
    --eval_steps 4000 \
    --save_steps 2000 \
    --model_name_or_path roberta-large \
    --output_dir ./result/WN18RR_roberta-large \
    --num_worker 12 \
    --seed 42 \
    --cls_method cls \
    --distance_metric euclidean \

CUDA_VISIBLE_DEVICES=2 \
python run_link_prediction.py \
    --model_class bert \
    --weight_decay 0.01 \
    --learning_rate 5e-5 \
    --adam_betas 0.9,0.98 \
    --adam_epsilon 1e-6 \
    --max_grad_norm 0. \
    --warmup_proportion 0.05 \
    --do_train --do_eval \
    --do_prediction \
    --num_train_epochs 7 \
    --dataset WN18RR \
    --max_seq_length 128 \
    --gradient_accumulation_steps 4 \
    --train_batch_size 16 \
    --eval_batch_size 128 \
    --logging_steps 100 \
    --eval_steps 4000 \
    --save_steps 2000 \
    --model_name_or_path bert-base-uncased \
    --do_lower_case \
    --output_dir ./result/WN18RR_bert \
    --num_worker 12 \
    --seed 42 \
    --cls_method cls \
    --distance_metric euclidean \

4.2 FB15k-237

CUDA_VISIBLE_DEVICES=0 \
python run_link_prediction.py \
    --model_class roberta \
    --weight_decay 0.01 \
    --learning_rate 1e-5 \
    --adam_betas 0.9,0.98 \
    --adam_epsilon 1e-6 \
    --max_grad_norm 0. \
    --warmup_proportion 0.05 \
    --do_train --do_eval \
    --do_prediction \
    --num_train_epochs 7. \
    --dataset FB15k-237 \
    --max_seq_length 100 \
    --gradient_accumulation_steps 4 \
    --train_batch_size 16 \
    --eval_batch_size 128 \
    --logging_steps 100 \
    --eval_steps -1 \
    --save_steps 2000 \
    --model_name_or_path roberta-large \
    --output_dir ./result/FB15k-237_roberta-large \
    --num_worker 12 \
    --seed 42 \
    --fp16 \
    --cls_method cls \
    --distance_metric euclidean \

4.3 UMLS

CUDA_VISIBLE_DEVICES=0 \
python run_link_prediction.py \
    --model_class roberta \
    --weight_decay 0.01 \
    --learning_rate 1e-5 \
    --adam_betas 0.9,0.98 \
    --adam_epsilon 1e-6 \
    --max_grad_norm 0. \
    --warmup_proportion 0.05 \
    --do_train --do_eval \
    --do_prediction \
    --num_train_epochs 20 \
    --dataset UMLS \
    --max_seq_length 16 \
    --gradient_accumulation_steps 1 \
    --train_batch_size 16 \
    --eval_batch_size 128 \
    --logging_steps 100 \
    --eval_steps -1 \
    --save_steps 200 \
    --model_name_or_path roberta-large \
    --output_dir ./result/UMLS_model \
    --num_worker 12 \
    --seed 42 \
    --cls_method cls \
    --distance_metric euclidean

4.4 NELL-One

CUDA_VISIBLE_DEVICES=0 \
python run_link_prediction.py \
    --model_class bert \
    --do_train --do_eval \usepacka--do_prediction \
    --warmup_proportion 0.1 \
    --learning_rate 5e-5 \
    --num_train_epochs 8. \
    --dataset NELL_standard \
    --max_seq_length 32 \
    --gradient_accumulation_steps 1 \
    --train_batch_size 16 \
    --eval_batch_size 128 \
    --logging_steps 100 \
    --eval_steps -1 \
    --save_steps 2000 \
    --model_name_or_path bert-base-uncased \
    --do_lower_case \
    --output_dir ./result/NELL_model \
    --num_worker 12 \
    --seed 42 \
    --fp16 \
    --cls_method cls \
    --distance_metric euclidean

5. StAR_Self-Adp

5.1 Data preprocessing

  • Get the trained model of RotatE, more details please refer to RotatE.

  • Run the below commands sequentially to get the training dataset of StAR_Self-Adp.

    • Run the run_get_ensemble_data.py in ./StAR
     CUDA_VISIBLE_DEVICES=0 python run_get_ensemble_data.py \
 	--dataset WN18RR \
 	--model_class roberta \
 	--model_name_or_path ./result/WN18RR_roberta-large \
 	--output_dir ./result/WN18RR_roberta-large \
 	--seed 42 \
 	--fp16
    • Run the ./codes/run.py in rotate. (please replace the TRAINED_MODEL_PATH with your own trained model's path)
     CUDA_VISIBLE_DEVICES=3 python ./codes/run.py \
 	--cuda --init ./models/RotatE_wn18rr_0 \
 	--test_batch_size 16 \
 	--star_info_path /home/wangbo/workspace/StAR_KGC-master/StAR/result/WN18RR_roberta-large \
 	--get_scores --get_model_dataset

5.2 Train and Test

  • Run the run.py in ./StAR/ensemble. Note the --mode should be alternate in head and tail, and perform a average operation to get the final results.
  • Note: Please replace YOUR_OUTPUT_DIR, TRAINED_MODEL_PATH and StAR_FILE_PATH in ./StAR/peach/common.py with your own paths to run the command and code.
CUDA_VISIBLE_DEVICES=2 python run.py \
--do_train --do_eval --do_prediction --seen_feature \
--mode tail \
--learning_rate 1e-3 \
--feature_method mix \
--neg_times 5 \
--num_train_epochs 3 \
--hinge_loss_margin 0.6 \
--train_batch_size 32 \
--test_batch_size 64 \
--logging_steps 100 \
--save_steps 2000 \
--eval_steps -1 \
--warmup_proportion 0 \
--output_dir /home/wangbo/workspace/StAR_KGC-master/StAR/result/WN18RR_roberta-large_ensemble  \
--dataset_dir /home/wangbo/workspace/StAR_KGC-master/StAR/result/WN18RR_roberta-large \
--context_score_path /home/wangbo/workspace/StAR_KGC-master/StAR/result/WN18RR_roberta-large \
--translation_score_path /home/wangbo/workspace/StAR_KGC-master/rotate/models/RotatE_wn18rr_0  \
--seed 42
Owner
Bo Wang
Ph.D. student at the School of Artificial Intelligence, Jilin University.
Bo Wang
GitHub Repository
AdaMML: Adaptive Multi-Modal Learning for Efficient Video Recognition

AdaMML: Adaptive Multi-Modal Learning for Efficient Video Recognition [ArXiv] [Project Page] This repository is the official implementation of AdaMML:

International Business Machines 43 Dec 26, 2022
A new framework, collaborative cascade prediction based on graph neural networks (CCasGNN) to jointly utilize the structural characteristics, sequence features, and user profiles.

CCasGNN A new framework, collaborative cascade prediction based on graph neural networks (CCasGNN) to jointly utilize the structural characteristics,

5 Apr 29, 2022
Workshop Materials Delivered on 28/02/2022

intro-to-cnn-p1 Repo for hosting workshop materials delivered on 28/02/2022 Questions you will answer in this workshop Learning Objectives What are co

Beginners Machine Learning 5 Feb 28, 2022
Unsupervised Foreground Extraction via Deep Region Competition

Unsupervised Foreground Extraction via Deep Region Competition [Paper] [Code] The official code repository for NeurIPS 2021 paper "Unsupervised Foregr

28 Nov 06, 2022
TPH-YOLOv5: Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-Captured Scenarios

TPH-YOLOv5 This repo is the implementation of "TPH-YOLOv5: Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-Captured

cv516Buaa 439 Dec 22, 2022
Project code for weakly supervised 3D object detectors using wide-baseline multi-view traffic camera data: WIBAM.

WIBAM (Work in progress) Weakly Supervised Training of Monocular 3D Object Detectors Using Wide Baseline Multi-view Traffic Camera Data 3D object dete

Matthew Howe 10 Aug 24, 2022
Network Enhancement implementation in pytorch

network_enahncement_pytorch Network Enhancement implementation in pytorch Research paper Network Enhancement: a general method to denoise weighted bio

Yen 1 Nov 12, 2021
DC540 hacking challenge 0x00005a.

dc540-0x00005a DC540 hacking challenge 0x00005a. PROMOTIONAL VIDEO - WATCH NOW HERE ON YOUTUBE CRITICAL PART 5A VIDEO - WATCH NOW HERE ON YOUTUBE Prio

Kevin Thomas 3 May 09, 2022
PyTorch implementation of our ICCV paper DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection.

Introduction This repo contains the official PyTorch implementation of our ICCV paper DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection. Up

133 Dec 29, 2022
This application explain how we can easily integrate Deepface framework with Python Django application

deepface_suite This application explain how we can easily integrate Deepface framework with Python Django application install redis cache install requ

Mohamed Naji Aboo 3 Apr 18, 2022
Open standard for machine learning interoperability

Open Neural Network Exchange (ONNX) is an open ecosystem that empowers AI developers to choose the right tools as their project evolves. ONNX provides

Open Neural Network Exchange 13.9k Dec 30, 2022
A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation

Segnet is deep fully convolutional neural network architecture for semantic pixel-wise segmentation. This is implementation of http://arxiv.org/pdf/15

Pradyumna Reddy Chinthala 190 Dec 15, 2022
This is an official implementation for "PlaneRecNet".

PlaneRecNet This is an official implementation for PlaneRecNet: A multi-task convolutional neural network provides instance segmentation for piece-wis

yaxu 50 Nov 17, 2022
Object-aware Contrastive Learning for Debiased Scene Representation

Object-aware Contrastive Learning Official PyTorch implementation of "Object-aware Contrastive Learning for Debiased Scene Representation" by Sangwoo

43 Dec 14, 2022
HAT: Hierarchical Aggregation Transformers for Person Re-identification

HAT: Hierarchical Aggregation Transformers for Person Re-identification

11 Sep 05, 2022
CUP-DNN is a deep neural network model used to predict tissues of origin for cancers of unknown of primary.

CUP-DNN CUP-DNN is a deep neural network model used to predict tissues of origin for cancers of unknown of primary. The model was trained on the expre

1 Oct 27, 2021
mbrl-lib is a toolbox for facilitating development of Model-Based Reinforcement Learning algorithms.

mbrl-lib is a toolbox for facilitating development of Model-Based Reinforcement Learning algorithms. It provides easily interchangeable modeling and planning components, and a set of utility function

Facebook Research 724 Jan 04, 2023
Experimental solutions to selected exercises from the book [Advances in Financial Machine Learning by Marcos Lopez De Prado]

Advances in Financial Machine Learning Exercises Experimental solutions to selected exercises from the book Advances in Financial Machine Learning by

Brian 1.4k Jan 04, 2023
Incremental Transformer Structure Enhanced Image Inpainting with Masking Positional Encoding (CVPR2022)

Incremental Transformer Structure Enhanced Image Inpainting with Masking Positional Encoding by Qiaole Dong*, Chenjie Cao*, Yanwei Fu Paper and Supple

Qiaole Dong 190 Dec 27, 2022
An Ensemble of CNN (Python 3.5.1 Tensorflow 1.3 numpy 1.13)

An Ensemble of CNN (Python 3.5.1 Tensorflow 1.3 numpy 1.13)

0 May 06, 2022