sdpa: support attn_mask.requires_grad, support expanded number of hea… #5332
| Job | Run time |
|---|---|
| 11m 32s | |
| 30m 48s | |
| 10m 52s | |
| 30m 31s | |
| 7m 52s | |
| 31m 31s | |
| 12m 49s | |
| 24m 12s | |
| 23m 37s | |
| 25m 26s | |
| 24m 14s | |
| 28m 6s | |
| 1s | |
| 4h 21m 31s |
| Job | Run time |
|---|---|
| 11m 32s | |
| 30m 48s | |
| 10m 52s | |
| 30m 31s | |
| 7m 52s | |
| 31m 31s | |
| 12m 49s | |
| 24m 12s | |
| 23m 37s | |
| 25m 26s | |
| 24m 14s | |
| 28m 6s | |
| 1s | |
| 4h 21m 31s |