Feat (nn/sdpa): quantization of scaled dot-product attention #1765
Job | Run time |
---|---|
3m 57s | |
3m 16s | |
4m 3s | |
3m 38s | |
4m 16s | |
3m 28s | |
5m 3s | |
4m 3s | |
4m 8s | |
3m 7s | |
4m 15s | |
3m 15s | |
2h 26m 14s | |
1h 15m 30s | |
4m 6s | |
3m 15s | |
4m 5s | |
3m 24s | |
4m 22s | |
3m 50s | |
5m 14s | |
4m 25s | |
4m 13s | |
3m 13s | |
4m 2s | |
3m 24s | |
2h 21m 4s | |
1h 23m 46s | |
9h 0m 36s |