Add head_dim = 64 in B200 Attention. (#4935) #18870
Job | Run time |
---|---|
10m 0s | |
12m 28s | |
16m 12s | |
9m 32s | |
14m 43s | |
25m 57s | |
26m 9s | |
26m 1s | |
26m 23s | |
1h 22m 48s | |
12m 30s | |
1h 24m 6s | |
33m 14s | |
20m 0s | |
33m 1s | |
20m 17s | |
7h 33m 21s |
Job | Run time |
---|---|
10m 0s | |
12m 28s | |
16m 12s | |
9m 32s | |
14m 43s | |
25m 57s | |
26m 9s | |
26m 1s | |
26m 23s | |
1h 22m 48s | |
12m 30s | |
1h 24m 6s | |
33m 14s | |
20m 0s | |
33m 1s | |
20m 17s | |
7h 33m 21s |