Feasibility of training on two A6000

I have two 48GB A6000 GPUs. Is it feasible to explore the training of this model on NuScenes with this configuration? What are the parameters that I would need to change to make it run on such a compute budget. The original paper talks about 8 A100 GPUs. 