bdvllrs/generalization-vision-pretrained

Pretrained models of Devillers, B., Choksi, B., Bielawski, R., & VanRullen, R. (2021). Does language help generalization in vision models? CoNLL 2021: Computational Natural Language Learning.

Pretrained models associated with the project: https://github.com/bdvllrs/generalization-vision.

To load the model using gensim (tried with gensim version 4.0.1 on python 3.8.5):

import gensim

model_path = "models/none_epoch-4.model"
model = gensim.models.Word2Vec.load(model_path)

The "none" model corresponds to a vanilla SkipGram trained without visual word embeddings (our baseline in the paper).