Inference Performance (FPS, Latency) on R2R
4.68FPSA-MMR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| A-MMRPruning Ratio=90%, Hardware=RTX 4090 GPU2026.03 | 4.68 | 5.61 | 213.4 | |
| SparseVLMPruning Ratio=90%, Hardware=RTX 4090 GPU2026.03 | 4.55 | 5.61 | 219.49 | |
| DivPrunePruning Ratio=90%, Hardware=RTX 4090 GPU2026.03 | 4.53 | 1.42 | 220.71 | |
| VisPrunerPruning Ratio=90%, Hardware=RTX 4090 GPU2026.03 | 4.46 | 9.86 | 224.36 | |
| UnprunedPruning Ratio=0%, Hardware=RTX 4090 GPU2026.03 | 4.32 | 10.94 | 231.34 |