Feat (nn/sdpa): quantization of scaled dot-product attention #1781
Job | Run time |
---|---|
4m 18s | |
3m 28s | |
5m 6s | |
4m 21s | |
3m 59s | |
3m 18s | |
3m 58s | |
3m 25s | |
2h 23m 32s | |
1h 20m 31s | |
2h 20m 21s | |
1h 25m 15s | |
2h 14m 18s | |
1h 24m 32s | |
4m 9s | |
3m 45s | |
5m 0s | |
4m 23s | |
4m 2s | |
3m 13s | |
4m 15s | |
3m 14s | |
2h 22m 20s | |
1h 18m 42s | |
2h 22m 3s | |
1h 23m 13s | |
2h 13m 9s | |
1h 24m 15s | |
23h 16m 5s |