llama-quant : correct n_attention_wv usage (#20357)
#46
| Job | Run time |
|---|---|
| 10m 41s | |
| 1s | |
| 2m 58s | |
| 3m 45s | |
| 2m 52s | |
| 8m 14s | |
| 2m 2s | |
| 7m 4s | |
| 54s | |
| 1h 26m 31s | |
| 2m 1s | |
| 27m 34s | |
| 11m 58s | |
| 2m 34s | |
| 5m 34s | |
| 4m 29s | |
| 17m 30s | |
| 10m 6s | |
| 15m 40s | |
| 0s | |
| 3h 42m 28s |