1 briefs tagged #speculative decoding.
The checkpoints add speculative decoding for three LFM2.5 models with llama.cpp and SGLang support.