[feat] CrossEncoder Training refactor - MultiGPU, loss logging, bf16, etc.
#1379
| Job | Run time |
|---|---|
| 14m 40s | |
| 19m 18s | |
| 14m 53s | |
| 17m 21s | |
| 16m 27s | |
| 16m 1s | |
| 16m 48s | |
| 20m 42s | |
| 2h 16m 10s |