cd scripts CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7' torchrun --master_port 15003 --nproc_per_node=8 \ train_triplane_diffusion.py \ --configs ../configs/finetune_triplane_diffusion.yaml \ --accum_iter 2 \ --output_dir ../output/finetune_dm/lowres_chair \ --log_dir ../output/finetune_dm/lowres_chair --num_workers 8 \ --batch_size 22 \ --blr 1e-4 \ --epochs 500 \ --dist_eval \ --warmup_epochs 20 \ --ae-pth ../output/ae/chair/best-checkpoint.pth \ --category chair \ --finetune \ --finetune-pth ../output/dm/chair/best-checkpoint.pth \ --data-pth ../data \ --replica 5