strata serve: prompt 4022 tokens = 0 reused + 4022 read in 12840 ms (313.2 tok/s), 256 generated in 22083 ms (11.6 tok/s), drafts accepted 164 of 220, 1 checkpoints
strata serve: decode expert cache hit rate: 76.1% (97064 hits / 127518 lookups)
strata serve: KV streaming: 99.35% of 1971840 block reads hit VRAM, 51.3 MiB read from RAM
strata serve: suffix drafts: 3 windows, 8 of 9 drafts accepted
