Attention Operator Performance on MLA 128 Head Dimension Sequence Length 16k
175.9TFLOPSLLM-TL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLM-TLHead Dimension=128, Mask=Causal, GPU=A100, Base Model=DeepSeek-V32025.06 | 175.9 | 9.41 | |
| cuDNNHead Dimension=128, Mask=Causal, GPU=A1002025.06 | 81.7 | — | |
| torchHead Dimension=128, Mask=Causal, GPU=A1002025.06 | 35.1 | — | |
| DeepSeek-V3Head Dimension=128, Mask=Causal, GPU=A1002025.06 | 18.7 | — |