Packed QKV and Rotary Embedding Support for sm<80 GQA#20012
Merged
YUNQIUGUO merged 4 commits intomainfrom aciddelgado/fix_rotary_memeff_gqaMar 23, 2024
+216-64
Commits
Commits on Mar 21, 2024
- committed
- committed
Commits on Mar 22, 2024
- committed
- committed