method · first seen 15 Oct 2026
activation compression
Also: activation checkpointing
Reducing memory used by stored intermediate activations during neural network training.
Mentions per week
PapersCodeDiscussion
Earliest sources
- BitSparse: Compact Activation Storage for ReLU Feed-Forward NetworksPapers · 15 Oct 2026 · Max Zhu, Bohang Sun, Pietro Lio, Jialin Yu
- PlaneZip: Lossless Bit-Plane Compression of Training Activations Offloaded over PCIePapers · 18 Oct 2026 · Zhufan Gu, Qizhe Tang, Yuxin Wang, Xun Zhou, Hui Feng
- Unlocking Turing Tensor Cores: A Measured Throughput Account of From-Scratch Pre-training on a Single 22G RTX 2080 TiPapers · 2 Nov 2026 · Dezhou Shen
- Know What You Need: Efficient Activation CheckpointingPapers · 6 Nov 2026 · Danyal Ahmed, Edouard Oyallon
Latest sources
- Know What You Need: Efficient Activation CheckpointingPapers · 6 Nov 2026 · Danyal Ahmed, Edouard Oyallon
- Unlocking Turing Tensor Cores: A Measured Throughput Account of From-Scratch Pre-training on a Single 22G RTX 2080 TiPapers · 2 Nov 2026 · Dezhou Shen
- PlaneZip: Lossless Bit-Plane Compression of Training Activations Offloaded over PCIePapers · 18 Oct 2026 · Zhufan Gu, Qizhe Tang, Yuxin Wang, Xun Zhou, Hui Feng
- BitSparse: Compact Activation Storage for ReLU Feed-Forward NetworksPapers · 15 Oct 2026 · Max Zhu, Bohang Sun, Pietro Lio, Jialin Yu