ReLU Attention on Llama2-7b (q=32, k=32) (1k)
150.88TFLOPSCuBridge
Evaluation Results
| Method | Links | |
|---|---|---|
| CuBridgeImplementation=CuBridge2026.05 | 150.88 | |
| Qimeng AttnImplementation=Qimeng Attention2026.05 | 88.56 | |
| TorchImplementation=Standard PyTorch2026.05 | 31.16 |