Repository navigation
hf_static v2: transposed key caches, full-length sliding caches, per-head attention #58286
| Job | Run time |
|---|---|
| 12s | |
| 7m 48s | |
| 17m 10s | |
| 9m 36s | |
| 9m 14s | |
| 6m 54s | |
| 8m 26s | |
| 12m 8s | |
| 9m 26s | |
| 14m 3s | |
| 4m 49s | |
| 14m 14s | |
| 14m 28s | |
| 6m 46s | |
| 15m 10s | |
| 14m 28s | |
| 2h 44m 52s |