Follow
Recycle-and-Distill: Universal Compression Strategy for Transformer-based Speech SSL Models with Attention Map Reusing and Masking Distillation. (arXiv:2305.11685v2 [eess.AS] UPDATED)
Recycle-and-Distill: Universal Compression Strategy for Transformer-based Speech SSL Models with Attention Map Reusing and Masking Distillation. (arXiv:2305.11685v2 [eess.AS] UPDATED)