Repository navigation
hf_static v2: transposed key caches, full-length sliding caches, per-head attention #1544
| Job | Run time |
|---|---|
| 31s | |
| 40m 7s | |
| 44m 6s | |
| 27m 54s | |
| 42m 29s | |
| 33m 26s | |
| 38m 8s | |
| 28m 43s | |
| 26m 49s | |
| 1h 4m 17s | |
| 5h 46m 30s |
| Job | Run time |
|---|---|
| 31s | |
| 40m 7s | |
| 44m 6s | |
| 27m 54s | |
| 42m 29s | |
| 33m 26s | |
| 38m 8s | |
| 28m 43s | |
| 26m 49s | |
| 1h 4m 17s | |
| 5h 46m 30s |