List your company

Prefill Latency on Llama-3-8B 500 Mbps

1.485Prefill Latency (s)BP
1.13883.475655.81258.14935May 25, 2025

Evaluation results

MethodLinks
BPStrategy=Block Parallelism, Nb=42025.051.485
ASTRAG=12025.051.54
ASTRAG=162025.051.548
BPStrategy=Block Parallelism, Nb=82025.051.561
ASTRAG=322025.051.583
SPStrategy=Sequence Parallelism2025.052.052
Llama-3-8B (Single-device)Device=Single-device2025.054.578
TPStrategy=Tensor Parallelism2025.0510.14