- python3
- pandas
- gensim
- numpy
- torchvision
- pytorch 1.11.0
- scipy
- transformers
- numpy
- tqdm
- ffmpeg
- h5py
- decord
- opencv-python
- Pillow
- Download and put the dataset in the
gsmt_data
folder: https://tinyurl.com/gsmt15042024 - Train the model by running
bash ./shells/train/{dataset}_gsmt.sh
(egoqa or madqa) - Evaluate the model via executing
bash ./shells/test/{dataset}_gsmt.sh
(egoqa or madqa)