Research map / Architectures and efficiency / RNNs and normalization
Coop: Memory is not a Commodity
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Coop instead of DTE | BERT Large · BERT Large, 50% memory ratio | compute overhead: 29% -> 11% | Helped |
| use Coop instead of DTR | BERT Large · BERT Large, 50% memory ratio | compute overhead: 41% -> 11% | Helped |
| co-optimize tensor rematerialization and allocation using Coop | GPT-3 style 2.7B · GPT-3 style 2.7B | minimum memory budget: 25% | Helped |
| use Coop's sliding-window eviction and allocation approach | all eight DNNs · all eight DNNs | memory fragmentation rate: less than 5% | Helped |
| remove any one of sliding window, recomputable in-place, or cheap tensor partitioning (ablation) | Coop · ResNet-50 | lowest supported memory ratio: 30% -> 40% | Hurt |
| replace sliding-window search with the DTE heuristic (ablation) | Coop · U-Net, 40% memory ratio | compute overhead: doubles the overhead | Hurt |
| remove recomputable in-place (ablation) | Coop · averaged across evaluated DNNs | memory fragmentation rate | Hurt |
| remove sliding-window eviction (ablation) | Coop · averaged across evaluated DNNs | memory fragmentation rate | Hurt |
| remove cheap tensor partitioning (ablation) | Coop · averaged across evaluated DNNs | memory fragmentation rate | No clear effect |
| use Coop rather than DTR or DTE | eight evaluated DNNs · all eight DNNs | compute overhead | Helped |
| use Coop rather than DTR or DTE | eight evaluated DNNs · all eight DNNs | search latency | Helped |
Closest papers by meaning
- HiRemate: Hierarchical Approach for Efficient Re-materialization of Neural Networks (ICML 2025 · similarity 0.87)
- COGNATE: Acceleration of Sparse Tensor Programs on Emerging Hardware using Transfer Learning (ICML 2025 · similarity 0.85)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · similarity 0.85)
- SMMF: Square-Matricized Momentum Factorization for Memory-Efficient Optimization (AAAI 2025 · similarity 0.85)
- Tensorion: A Tensor-Aware Generalization of the Muon Optimizer (NeurIPS 2026 · similarity 0.84)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.