TokenWeave achieves up to 1.28x lower latency and 1.19x higher throughput for distributed LLM inference by enabling compute-communication overlap at small token counts via a fused AllReduce-RMSNorm kernel that uses only 2-8 SMs.
D ARTIFACT D.1 Abstract Distributed inference of LLMs can incur overheads of up to 20%, even when GPUs are connected via high-speed interconnects such as NVLink
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TokenWeave: Efficient Compute-Communication Overlap for Distributed LLM Inference
TokenWeave achieves up to 1.28x lower latency and 1.19x higher throughput for distributed LLM inference by enabling compute-communication overlap at small token counts via a fused AllReduce-RMSNorm kernel that uses only 2-8 SMs.