Repository navigation
hf_static v2: transposed key caches, full-length sliding caches, per-head attention #1379
| Job | Run time |
|---|---|
| 9s | |
| 16m 5s | |
| 27m 11s | |
| 23m 7s | |
| 29m 42s | |
| 23m 24s | |
| 25m 4s | |
| 2h 24m 42s |
| Job | Run time |
|---|---|
| 9s | |
| 16m 5s | |
| 27m 11s | |
| 23m 7s | |
| 29m 42s | |
| 23m 24s | |
| 25m 4s | |
| 2h 24m 42s |