Vilio: State-of-the-art Visio-Linguistic Models applied to Hateful Memes
This work provides a user-friendly starting point for researchers and practitioners working on visio-linguistic problems, demonstrating strong performance on the Hateful Memes Challenge.
This work presents Vilio, an implementation of state-of-the-art visio-linguistic models applied to the Hateful Memes Dataset. An ensemble of 5 different V+L models implemented in Vilio achieved 2nd place out of 3,300 participants in the Hateful Memes Challenge.
This work presents Vilio, an implementation of state-of-the-art visio-linguistic models and their application to the Hateful Memes Dataset. The implemented models have been fitted into a uniform code-base and altered to yield better performance. The goal of Vilio is to provide a user-friendly starting point for any visio-linguistic problem. An ensemble of 5 different V+L models implemented in Vilio achieves 2nd place in the Hateful Memes Challenge out of 3,300 participants. The code is available at https://github.com/Muennighoff/vilio.