#加载预训练的模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=6 python3 -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
checkpointing.pretrained_model=pretrained/depthsplat-gs-base-re10k-256x448-view2-76a0605a.pth \
wandb.project=depthsplat \
model.encoder.supervise_intermediate_depth=false \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6


#不加载预训练的模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=0 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
wandb.project=depthsplat \
model.encoder.supervise_intermediate_depth=false \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6


#测试修改的模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=0 python -m src.test_model +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["/mnt/pfs/users/wangweijie/yeqing/BEV-Splat/datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
wandb.project=depthsplat \
model.encoder.supervise_intermediate_depth=false \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6


##训练修改之后的模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=0 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=6 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
wandb.project=VolSplat \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6


#测试depthanything模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=6 python -m src.test.try_depthanything


##测试depthanything的深度
conda activate depthsplat
CUDA_VISIBLE_DEVICES=5 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=6 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
model.encoder.supervise_intermediate_depth=false \
wandb.project=VolSplat \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6



##测试volSplat模型
conda activate depthsplat
CUDA_VISIBLE_DEVICES=4 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
model.encoder.supervise_intermediate_depth=false \
wandb.project=VolSplat \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6 \
train.no_viz_video=True


##测试volSplat模型（6-10张图片）
conda activate depthsplat
CUDA_VISIBLE_DEVICES=6 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=5 \
dataset.max_views=10 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
model.encoder.supervise_intermediate_depth=false \
wandb.project=VolSplat \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6 \
train.no_viz_video=True

 
##训练volSplat模型（加载预训练的depthanythingencoder） #
conda activate /mnt/pfs/users/chaojun.ni/wangweijie_mnt/wangweijie/miniconda3/envs/depthsplat
CUDA_VISIBLE_DEVICES=2 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
checkpointing.pretrained_monodepth=pretrained/pretrained_weights/depth_anything_v2_vitb.pth \
checkpointing.pretrained_mvdepth=pretrained/pretrained_weights/gmflow-scale1-things-e9887eda.pth \
wandb.project=VolSplat-Unet \
model.encoder.supervise_intermediate_depth=True \
output_dir=outputs/dl3dv-256x448-volsplat_3dunet \
train.no_viz_video=True

##重新加载模型训练
conda activate /mnt/pfs/users/chaojun.ni/wangweijie_mnt/wangweijie/miniconda3/envs/depthsplat
CUDA_VISIBLE_DEVICES=2 python -m src.main +experiment=dl3dv \
data_loader.train.batch_size=1 \
'dataset.roots'='["datasets/dl3dv"]' \
dataset.view_sampler.num_target_views=8 \
dataset.view_sampler.num_context_views=6 \
dataset.min_views=1 \
dataset.max_views=6 \
trainer.max_steps=100000 \
trainer.num_nodes=1 \
model.encoder.num_scales=2 \
model.encoder.upsample_factor=4 \
model.encoder.lowest_feature_resolution=8 \
model.encoder.monodepth_vit_type=vitb \
checkpointing.resume=True \
wandb.project=VolSplat-Unet \
output_dir=checkpoints/dl3dv-256x448-depthsplat-base-randview2-6 \
model.encoder.supervise_intermediate_depth=True \
train.no_viz_video=True