Hello, I have a question. I saw that the LAMP algorithm was implemented in the code of the paper, but the description in the original paper said that LAMP is an unstructured pruning algorithm, but why is it structured in your library? Appears in the form of pruning, sincerely hope to get your answer
Hi @Kk875, LAMP is a normalization method actually. It normalizes importance scores from different layers into a comparable scale. So, it's easy and natural to adapt LAMP to structured pruning.
Hello, I have a question. I saw that the LAMP algorithm was implemented in the code of the paper, but the description in the original paper said that LAMP is an unstructured pruning algorithm, but why is it structured in your library? Appears in the form of pruning, sincerely hope to get your answer