Cohere For AI - Guest Speaker: Lucas Beyer,Researcher @Google DeepMind

About Speaker: Bio: Lucas grew up in Belgium wanting to make video games and their AI, went on to study mechanical engineering at RWTH Aachen in Germany, did a PhD in robotic perception/computer vision there too, and is now researching representation learning and vision backbones at Google DeepMind/Brain in Zürich. Website: http://lucasb.eyer.be/
Talk Details: We propose a simple pairwise Sigmoid loss for Language-Image Pre-training (SigLIP). Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. Combined with Locked-image Tuning, with only four TPUv4 chips, we train a SigLiT model that achieves 84.5% ImageNet zero-shot accuracy in two days. The disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. abs: https://arxiv.org/abs/2303.15343

Add event to calendar

Apple Google Office 365 Outlook Outlook.com Yahoo