[core / DDP] Fix RM trainer + DDP + quantization + propagate gradient_checkpointing_kwargs in SFT & DPO
#211
| Job | Run time |
|---|---|
| 3s | |
| 3s |