model support issues

I am trying to train on Qwen/Qwen3, but the supporting transformer version higher than 4.51. And verl 0.1 requires transformers<4.48, but you have transformers 4.51.0 which is incompatible.

Can you fix it to fit into the new models?