LoRA support?

[soumyasj]

Dear Authors,

Thank you for providing the code for this work! I wanted to know if it supports LoRA finetuning with quantization.

[BambiSheng]

Hi, thanks for your interest in our work!
Actually we don't have any plans to support LoRA as it's outside of our scope, and the training behaviors may not meet expectations. Nevertheless you may have a try (verl's official tutorial for lora) since verl introduced its LoRA support after v0.4.0 and our latest code is based on v0.4.1.