@felurian8 @BasedBeffJezos latent space is the model's internal representation of what it's learned from all the training data, compressed via the training cycles to capture the most important features