Efficient Training of Sparse Autoencoders for Large Language Models via Layer Groups

Open in new window