Comment Comment by elcomet We didn't have the compute required (GPUs powerful enough to parallelize forward and backward pass). This compute is what allows us to train from human knowledge or distillation. parent elcomet 3 months ago