VILA is a visual language model (VLM) pre-trained with interleaved image-text data at scale, enabling multi-image VLM. VILA is deployable on the edge.
Differentiable augmentation to improve the data efficiency of GAN training.
