Is reinforcement learning supported by DeepSpeed?

[kisseternity]

Hello, these days reinforcement training in NLP is really popular. I wonder if I can use DeepSpeed in reinforcement training.
as the procedure of reinforcement learning is quite different from the normal training as demonstrated in official examples. (I tend to suppose it's not supported yet by DeepSpeed). Any idea is helpful, thanks.