Dust Trains AI Models Without Backpropagation

Original: Dust: Pretraining Transformers Without Backpropagation

Source: qlabs.sh ↗