Cutting inference cold starts by 40x with LP, FUSE, C/R, and CUDA-checkpointmodal.com
11 points by charles_irl 28 minutes ago | 0 comments