mirror of
https://github.com/likelovewant/ollama-for-amd.git
synced 2025-12-21 14:26:30 +00:00
In cases where we allocate a tensor and then fully overwrite it with copied data, it is wasteful to first zero out the memory.
In cases where we allocate a tensor and then fully overwrite it with copied data, it is wasteful to first zero out the memory.