config.py

# paths
qa_path = 'data'  # directory containing the question and annotation jsons
bottom_up_trainval_path = '/home/user/data/data/trainval_36'  # directory containing the .tsv file(s) with bottom up features
bottom_up_test_path = '/home/user/data/data/test2014_36'  # directory containing the .tsv file(s) with bottom up features
preprocessed_trainval_path = 'genome-trainval.h5'  # path where preprocessed features from the trainval split are saved to and loaded from
preprocessed_test_path = 'genome-test.h5'  # path where preprocessed features from the test split are saved to and loaded from
vocabulary_path = 'vocab.json'  # path where the used vocabularies for question and answers are saved to

task = 'OpenEnded'
dataset = 'mscoco'

test_split = 'test2014'  # either 'test-dev2015' or 'test2015'

# preprocess config
output_size = 36  # max number of object proposals per image
output_features = 2048  # number of features in each object proposal

# training config
epochs = 100
batch_size = 32 
initial_lr = 1.5e-4
lr_halflife = 50000  # in iterations
data_workers = 4
max_answers = 3000