Skip to content

Drop float32 weight copy from BitLinear to cut resident memory ~80% per linear layer#24

Merged
sharpninja merged 4 commits intomainfrom
copilot/improve-memory-performance
Mar 21, 2026
Merged

Drop float32 weight copy from BitLinear to cut resident memory ~80% per linear layer#24
sharpninja merged 4 commits intomainfrom
copilot/improve-memory-performance

Commits

Commits on Mar 21, 2026