Research map / Architectures and efficiency / Pruning and efficient tuning
CPrune: Compiler-Informed Model Pruning for Efficient Target-Aware DNN Execution
Taeho Kim, Yongin Kwon, Jemin Lee, Taeho Kim, Sangtae Ha
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add compiler-informed tuning to CPrune pruning (ablation) | CPrune · CIFAR-10 (ResNet-18, Kryo 585) | FPS: 57.77 -> 93.63 | Improved |
| prune all subgraphs associated with a task instead of a single subgraph (ablation) | CPrune · CIFAR-10 (ResNet-18, Kryo 585) | FPS: 79.62 -> 93.63 | Improved |
| prune all associated subgraphs instead of one subgraph (accuracy) (ablation) | CPrune · CIFAR-10 (ResNet-18, Kryo 585) | Top-1 accuracy: 94.27% -> 94.14% | No clear effect |
| use compiler-informed CPrune instead of NetAdapt+TVM | ResNet-18 · ImageNet (Kryo 385 CPU) | FPS: 35.17 -> 36.92 | Improved |
| use CPrune instead of FPGM+TVM | ResNet-18 · ImageNet (Mali-G72 GPU) | FPS: 26.62 -> 30.02 | Improved |
| use CPrune instead of AMC+TVM | MobileNetV2 · ImageNet (Kryo 385 CPU) | FPS: 67.62 -> 76.92 | Improved |
| use CPrune instead of AMC+TVM | MobileNetV2 · ImageNet (Mali-G72 GPU) | FPS: 90.58 -> 101.56 | Improved |
| apply CPrune on top of TVM auto-tune | ResNet-18 · CIFAR-10 (Kryo 280 CPU) | FPS: 33.82 -> 109.45 | Improved |
| apply CPrune on top of TVM auto-tune | MnasNet1.0 · ImageNet (Kryo 585 CPU) | FPS: 42.92 -> 56.85 | Improved |
| pruning (PQF) before TVM compilation instead of original model | ResNet-18 · ImageNet (Kryo 385 CPU) | FPS: 18.86 -> 18.73 | No clear effect |
| raise accuracy requirement above 89% so pruning stops near iteration 6 | CPrune · ImageNet (ResNet-18, Kryo 385) | FPS increase rate: 1.96 -> 1.3 times | Hurt |
| select the fastest pruned model before compilation instead of after compiler optimization | VGG-16 pruned models · CIFAR-10 (RTX-3080) | FPS: 2857 -> 205 | Hurt |
Later papers that compared with this method
2 later accepted papers report C-PRUNE as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- C-GNN-PRUNE: A Unified Graph-Based Framework for Structure-Aware Pruning of Mixture-of-Experts Models (AAAI 2026 · better on 6 of 6 shared results)
- Attribution-Guided and Coverage-Maximized Pruning for Structural MoE Compression (ICML 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- NPAS: A Compiler-Aware Framework of Unified Network Pruning and Architecture Search for Beyond Real-Time Mobile Acceleration (CVPR 2021 · similarity 0.87)
- DepthShrinker: A New Compression Paradigm Towards Boosting Real-Hardware Efficiency of Compact Neural Networks (ICML 2022 · similarity 0.86)
- PCONV: The Missing but Desirable Sparsity in DNN Weight Pruning for Real-Time Execution on Mobile Devices (AAAI 2020 · similarity 0.86)
- GRIM: A General, Real-Time Deep Learning Inference Framework for Mobile Devices Based on Fine-Grained Structured Weight Sparsity (TPAMI 2021 · similarity 0.86)
- Auto-Train-Once: Controller Network Guided Automatic Network Pruning from Scratch (CVPR 2024 · similarity 0.83)
- REPrune: Channel Pruning via Kernel Representative Selection (AAAI 2024 · similarity 0.83)
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.
