Search papers, labs, and topics across Lattice.
HIT(SZ)
2
0
5
KernelFlume slashes operational costs by up to 61% while maintaining low latency for long-context LLM decoding, revolutionizing how we scale model serving.
LLM training bottlenecks? ZipCCL achieves up to 1.18x end-to-end speedups by losslessly compressing communication collectives, without sacrificing model quality.
