Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference

no code implementations • 14 Mar 2024 • Piotr Nawrot, Adrian Łańcucki, Marcin Chochowski, David Tarjan, Edoardo M. Ponti

As a solution, we propose Dynamic Memory Compression (DMC), a method for on-line key-value cache compression at inference time.

SDCNet: Video Prediction Using Spatially-Displaced Convolution

1 code implementation • 2 Nov 2018 • Fitsum A. Reda, Guilin Liu, Kevin J. Shih, Robert Kirby, Jon Barker, David Tarjan, Andrew Tao, Bryan Catanzaro

We present an approach for high-resolution video frame prediction by conditioning on both past frames and past optical flows.

Optical Flow Estimation SSIM +1

SDC-Net: Video prediction using spatially-displaced convolution

1 code implementation • ECCV 2018 • Fitsum A. Reda, Guilin Liu, Kevin J. Shih, Robert Kirby, Jon Barker, David Tarjan, Andrew Tao, Bryan Catanzaro

We present an approach for high-resolution video frame prediction by conditioning on both past frames and past optical flows.

Optical Flow Estimation SSIM +1