1 paper
Elias Frantar, Carlos Riquelme, Neil Houlsby +2
We explore the impact of parameter sparsity on the scaling behavior of Transformers trained on massive datasets (i.e., "foundation models"), in both vision and language domains. In…