run.sh

# signle gpu
python \
    ./src/tasks/train.py \
    --config ./src/configs/favd_32frm_default.json \
    --per_gpu_train_batch_size 2 \
    --per_gpu_eval_batch_size 2 \
    --num_train_epochs 150 \
    --learning_rate 0.0001 \
    --max_num_frames 32 \
    --backbone_coef_lr 0.05 \
    --learn_mask_enabled \
    --loss_sparse_w 0.5 \
    --lambda_ 0.1 \
    --output_dir ./output/favd_default \

# multiple gpus
torchrun --nproc_per_node=8 \
    ./src/tasks/train.py \
    --config ./src/configs/favd_32frm_default.json \
    --per_gpu_train_batch_size 2 \
    --per_gpu_eval_batch_size 2 \
    --num_train_epochs 150 \
    --learning_rate 0.0001\
    --max_num_frames 32 \
    --backbone_coef_lr 0.05 \
    --learn_mask_enabled \
    --loss_sparse_w 0.5 \
    --lambda_ 0.1 \
    --output_dir ./output/favd_default \

# multiple nodes
torchrun --nproc_per_node=8 \
    --master_addr= \
    --master_port= \
    --nnodes= \
    --node_rank= \
    --config ./src/configs/favd_32frm_default.json \
    --per_gpu_train_batch_size 2 \
    --per_gpu_eval_batch_size 2 \
    --num_train_epochs 150 \
    --learning_rate 0.0001 \
    --max_num_frames 32 \
    --backbone_coef_lr 0.05 \
    --learn_mask_enabled \
    --loss_sparse_w 0.5 \
    --lambda_ 0.1 \
    --output_dir ./output/favd_default \