Vision Transformer Compression with Structured Pruning and Low Rank Approximation

Ankur Kumar

Transformer architecture has gained popularity due to its ability to scale with large dataset. Consequently, there is a need to reduce the model size and latency, especially for on…