Skip to content

fix: build_attn V-padded reshape uses Q-head count, not KV-head count (#78)#116

Merged
TheTom merged 1 commit into
feature/turboquant-kv-cachefrom
fix/issue-78-gqa-reshape
May 1, 2026
Merged

fix: build_attn V-padded reshape uses Q-head count, not KV-head count (#78)#116
TheTom merged 1 commit into
feature/turboquant-kv-cachefrom
fix/issue-78-gqa-reshape

Commits

Commits on May 1, 2026