Identifying relevant common sense information in knowledge graphs

Code for our paper Identifying relevant common sense information in knowledge graphs, built on top of the code from Pan et al..

Multi-Hop Inference Explanation Regeneration (TextGraphs-15)

This is the code for the first place in the textgraphs-15 competition

paper ：DeepBlueAI at TextGraphs 2021 Shared Task: Treating Multi-HopInference Explanation Regeneration as A Ranking Problem

requirement

conda create -n deepblue python=3.8.5 numpy matplotlib ipython
conda activate deepblue
conda install pytorch=1.6 cudatoolkit=10.1 -c pytorch
pip install pandas==1.2.3 transformers==4.5.1 sklearn

1. pytorch=1.6
2. transformers=4.5.1
3. pandas=1.2.3
4. cuda=10.1
5. python=3.8.5

pre-training model

roberta-large

https://huggingface.co/roberta-large/tree/main

ernie-2.0-large-en

https://huggingface.co/nghuyong/ernie-2.0-large-en/tree/main

run the code

recall train

python recall_trainer.py --output_dir=save_model/recall/roberta --bert_path=roberta-large --per_gpu_batch_size 48 

python recall_trainer.py --output_dir=save_model/recall/ernie --bert_path=nghuyong/ernie-2.0-large-en --per_gpu_batch_size 48 

recall predict

python recall_predict.py

sort train

python sort_trainer.py --output_dir=save_model/sort/roberta --bert_path=roberta-large --per_gpu_batch_size 48

python sort_trainer.py --output_dir=save_model/sort/ernie --bert_path=nghuyong/ernie-2.0-large-en --per_gpu_batch_size 48

sort predict

python sort_predict.py

result

The result is "result/predict.txt"