WIP: [GPU] Use FP32 accumulator for QK multiplication for 2nd+ token calculation in PagedAttention #17056
| Job | Run time |
|---|---|
| 21s | |
| 35m 33s | |
| 0s | |
| 0s | |
| 7m 58s | |
| 0s | |
| 0s | |
| 0s | |
| 57s | |
| 4m 21s | |
| 9m 47s | |
| 0s | |
| 1s | |
| 58m 58s |
| Job | Run time |
|---|---|
| 21s | |
| 35m 33s | |
| 0s | |
| 0s | |
| 7m 58s | |
| 0s | |
| 0s | |
| 0s | |
| 57s | |
| 4m 21s | |
| 9m 47s | |
| 0s | |
| 1s | |
| 58m 58s |