[Kernel][Attention] Separate Attention.kv_scale
into k_scale
and v_scale
#15204
Job | Run time |
---|---|
1m 13s | |
1m 12s | |
1m 11s | |
53s | |
4m 29s |
Attention.kv_scale
into k_scale
and v_scale
#15204
Job | Run time |
---|---|
1m 13s | |
1m 12s | |
1m 11s | |
53s | |
4m 29s |