Code and Experiments for ACL-IJCNLP 2021 Paper "Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering."
Fix the bugs that were reported in issue #2 and add reproducing environment on dokcer.
Bug
VQA's data folder name. change data/VQA-Images and data/VQA-Spatials to data/VQA2-Images and data/VQA2-Spatials respectively.
GQA's image path.
return value in training_epoch_end of each model.
arguments of ModelCheckpoint.
command for creating DatasetMap.
Add
Provide Docker environment for reproducing.
I added files, Dockerfile and requirements.txt.
And, I reported how to build image and run container in README.
Attention
I have tested the PR in Docker environment, but not on Anaconda.
Fix the bugs that were reported in issue #2 and add reproducing environment on dokcer.
Bug
data/VQA-Imagesanddata/VQA-Spatialstodata/VQA2-Imagesanddata/VQA2-Spatialsrespectively.training_epoch_endof each model.ModelCheckpoint.Add
Provide Docker environment for reproducing. I added files,
Dockerfileandrequirements.txt. And, I reported how to build image and run container inREADME.Attention
I have tested the PR in Docker environment, but not on Anaconda.