Skip to content

Latest commit

 

History

History
56 lines (46 loc) · 2.42 KB

TRAIN.md

File metadata and controls

56 lines (46 loc) · 2.42 KB
Raw

Train New Models

Pretraining

export MASTER_ADDR=$DIST_0_IP
export MASTER_PORT=$DIST_0_PORT
export NODE_RANK=$DIST_RANK
python run.py with data_root=<ARROW_ROOT> num_gpus=<NUM_GPUS> num_nodes=<NUM_NODES> task_mlm_itm whole_word_masking=True step200k per_gpu_batchsize=<BS_FITS_YOUR_GPU>

ex)
python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_mlm_itm whole_word_masking=True step200k per_gpu_batchsize=64

Finetune on VQAv2

export MASTER_ADDR=$DIST_0_IP
export MASTER_PORT=$DIST_0_PORT
export NODE_RANK=$DIST_RANK
python run.py with data_root=<ARROW_ROOT> num_gpus=<NUM_GPUS> num_nodes=<NUM_NODES> task_finetune_vqa_trainval_randaug per_gpu_batchsize=<BS_FITS_YOUR_GPU> load_path="<YOUR_WEIGHT_ROOT>/vilt_200k_mlm_itm.ckpt"

ex)
python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_finetune_vqa_trainval_randaug per_gpu_batchsize=64 load_path="weights/vilt_200k_mlm_itm.ckpt"

Finetune on NLVR2

export MASTER_ADDR=$DIST_0_IP
export MASTER_PORT=$DIST_0_PORT
export NODE_RANK=$DIST_RANK
python run.py with data_root=<ARROW_ROOT> num_gpus=<NUM_GPUS> num_nodes=<NUM_NODES> task_finetune_nlvr2_randaug per_gpu_batchsize=<BS_FITS_YOUR_GPU> load_path="<YOUR_WEIGHT_ROOT>/vilt_200k_mlm_itm.ckpt"

ex)
python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_finetune_nlvr2_randaug per_gpu_batchsize=32 load_path="weights/vilt_200k_mlm_itm.ckpt"

Finetune on COCO IR/TR

export MASTER_ADDR=$DIST_0_IP
export MASTER_PORT=$DIST_0_PORT
export NODE_RANK=$DIST_RANK
python run.py with data_root=<ARROW_ROOT> num_gpus=<NUM_GPUS> num_nodes=<NUM_NODES> task_finetune_irtr_coco_randaug per_gpu_batchsize=<BS_FITS_YOUR_GPU> load_path="<YOUR_WEIGHT_ROOT>/vilt_200k_mlm_itm.ckpt"

ex)
python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_finetune_irtr_coco_randaug per_gpu_batchsize=4 load_path="weights/vilt_200k_mlm_itm.ckpt"

Finetune on F30K IR/TR

export MASTER_ADDR=$DIST_0_IP
export MASTER_PORT=$DIST_0_PORT
export NODE_RANK=$DIST_RANK
python run.py with data_root=<ARROW_ROOT> num_gpus=<NUM_GPUS> num_nodes=<NUM_NODES> task_finetune_irtr_f30k_randaug per_gpu_batchsize=<BS_FITS_YOUR_GPU> load_path="<YOUR_WEIGHT_ROOT>/vilt_200k_mlm_itm.ckpt"

ex)
python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_finetune_irtr_f30k_randaug per_gpu_batchsize=4 load_path="weights/vilt_200k_mlm_itm.ckpt"