mirror of
https://github.com/ollama/ollama.git
synced 2026-07-23 09:10:53 -05:00
d573a2367b
The MTP validation forward schedules a snapshot at every drafted token, which made CausalConv1D re-run the depthwise conv once per segment to recover each boundary's conv tail. A conv boundary state is just the trailing convTail input positions, so run the conv once over the whole window and slice each boundary tail from the shared buffer, removing the per-token conv launches.