Rename fgqa_hs.py to TaxonomiGQA.py

Update fgqa_hs.py

Update fgqa_hs.py

update custom script

update custom script

updated to load images from tar file

Added custom loading script to better text-only and image-text cases

updated metadata file

updated metadata file

removed unused images no longer needed by metadata

updated metadata file

updated metadata file

updated metadata file

updated metadata file

updated metadata

Update: Simplified Setup

fixing bugs...

fixing bugs...

fixing bugs...

fixing bugs...

fixing bugs...

bug fix in dataset generator

Create README.md

Updated dataloader and files

Upload images.tar

Delete model_results

Delete val/single_nouns_balanced.json

Upload data.tsv

update sampled nouns, adding new questions, fix the bug

Rename val/sample_single_nouns.json to val/sampled_single_nouns.json

old version deleted

update dataset (before sampling)

update dataset (before sampling)

update dataset

old files deleted

old version deleted

upload CSV FILE

upload CSV FILE

upload CSV FILE

upload CSV FILE

upload single noun file with sampling

upload val single nouns data file after second round of filtering

Fixed loader questionType not currently available to pass on, and fixed filtering of samples to only match the chosen images.

Fixed loader questionType not currently available to pass on, and fixed filtering of samples to only match the chosen images.

Deleting GQA training data json files, not used

Fixed loader to keep original question type and renaming substitution hops.

Added a few missing images.

Tracking validation images locally instead, including images matching our filtering criteria that were missing from the sample data.

Update fgqa_hs.py

Upload single_nouns.jsonl