python3 eval.py --config configs/gnt_full.txt --eval_scenes orchids --expname gnt_full --chunk_size 500 --run_val --N_samples 192 python3 eval.py --config configs/gnt_llff.txt --eval_scenes orchids --expname gnt_llff --chunk_size 500 --run_val --N_samples 192

python3 eval.py --config configs/gnt_full.txt --eval_dataset rffr --eval_scenes art1 --expname gnt_full --chunk_size 500 --run_val --N_samples 192

export CUDA_VISIBLE_DEVICES=5 python3 eval.py --config configs/gnt_ft_rffr.txt --eval_dataset rffr --eval_scenes art1 --expname gen_ft_rffr --chunk_size 500 --run_val --N_rand 1024 --N_samples 192 --ckpt_path ./out/gnt_full/model_720000.pth

CUDA_VISIBLE_DEVICES=5 python3 eval.py --config configs/gnt_full.txt --eval_scenes room --expname gnt_room --run_val --N_rand 1024 --N_samples 192 --ckpt_path ./out/model_best.pth

python3 render.py --config configs/gnt_full.txt --eval_scenes orchids --expname gnt_full --chunk_size 500 --run_val --N_samples 192

export CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train.py --config configs/gnt_ft_rffr.txt --expname vanilla_Nray --N_rand 2048

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train.py --config configs/gnt_ft_rffr.txt --expname low_Nray --N_rand 1024

export CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train.py --config configs/gnt_ft_rffr.txt --expname vanilla_Nray --N_rand 1024 --N_samples 192

python -m torch.distributed.launch --nproc_per_node=8
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/cra/train_gnt_scannet.yaml

python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/cra/train_gnt_scannet.yaml
--ckpt_path ./out/gnt_full/model_best.pth --expname fc_gnt_pretrain --no_load_opt

export CUDA_VISIBLE_DEVICES=6,7 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/cra/train_gnt_scannet.yaml
--ckpt_path ./out/model_best.pth --expname LOW_LR

export CUDA_VISIBLE_DEVICES=4,5 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train.py --config configs/gnt_full.txt --expname gnt_full --N_rand 512 --N_importance 32

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/cra/train_gnt_scannet.yaml
--ckpt_path ./out/ibrnet_best.pth --expname ibrnet --model ibrnet --no_load_opt

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet.txt
--ckpt_path ./out/gnt_best.pth --expname gnt_semantic --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=4,5,6,7 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet.txt
--ckpt_path ./out/gnt_best.pth --expname gnt_smeantic_full --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet.txt
--ckpt_path ./out/gnt_best.pth --expname gnt_smeantic_train_val1 --val_set_list configs/scannetv2_val_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=4,5,6,7 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_lr3.txt
--ckpt_path ./out/gnt_best.pth --expname org_train --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_lr3.txt
--ckpt_path ./out/gnt_best.pth --expname lr3 --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_fuxian.txt
--ckpt_path ./out/gnt_best.pth --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_fuxian.txt
--ckpt_path ./out/gnt_best.pth --expname only_que --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=4,5 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de1.txt
--ckpt_path ./out/gnt_best.pth --expname only_que_de1 --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de2.txt
--ckpt_path ./out/gnt_best.pth --expname only_que_de2 --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de3.txt

export CUDA_VISIBLE_DEVICES=6,7 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_batch_scannet.py --config configs/gnt_scannet_batch.txt

export CUDA_VISIBLE_DEVICES=2,3 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_unbounded.txt
--ckpt_path ./out/gnt_best.pth --val_set_list configs/scannetv2_test_split.txt --no_load_opt --no_load_scheduler

export CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
ft_scannet.py --config configs/gnt_scannet_ft.txt --expname gnt_ft2

export CUDA_VISIBLE_DEVICES=6,7 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_dino.txt --expname gnt_scannet_dino2

export CUDA_VISIBLE_DEVICES=4,5 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de2.txt --expname selected_inds --selected_inds

export CUDA_VISIBLE_DEVICES=6,7 python -m torch.distributed.launch --nproc_per_node=2
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de2_scale0.5.txt

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.launch --nproc_per_node=8
--master_port=$(( RANDOM % 1000 + 50000 ))
train_scannet.py --config configs/gnt_scannet_de2.txt --expname de2_gpu8