CUDA: Faster Mixtral prompt processing #6962
| Job | Run time |
|---|---|
| 5m 45s | |
| 1m 38s | |
| 3m 22s | |
| 1m 34s | |
| 1m 48s | |
| 4m 30s | |
| 5m 52s | |
| 3m 28s | |
| 3m 50s | |
| 15m 43s | |
| 2m 8s | |
| 6m 20s | |
| 5m 30s | |
| 1m 15s | |
| 4m 20s | |
| 3m 5s | |
| 5m 10s | |
| 2m 56s | |
| 4m 48s | |
| 3m 11s | |
| 3m 0s | |
| 19m 45s | |
| 9m 7s | |
| 24m 31s | |
| 4m 45s | |
| 2m 47s | |
| 0s | |
| 2h 30m 8s |