[Kernel][Attention] Separate Attention.kv_scale
into k_scale
and v_scale
#8542
Job | Run time |
---|---|
21s | |
24s | |
21s | |
22s | |
1m 28s |
Attention.kv_scale
into k_scale
and v_scale
#8542
Job | Run time |
---|---|
21s | |
24s | |
21s | |
22s | |
1m 28s |