scripts/train_elia_refcocog_umd.py

import os
import random

port = random.randrange(2, 65534)

cmd='NCCL_IB_DISABLE=1 NCCL_DEBUG=VERSION NCCL_P2P_DISABLE=1 python -m torch.distributed.launch --nproc_per_node="8" --master_port="{:d}" train_elia.py --model="lavt" --dataset="refcocog" --splitBy "umd" --model_id="refcocog_umd" --batch-size="4" --lr="0.00005" --wd="1e-2" --swin_type="base" --pretrained_swin_weights="./pretrained_weights/swin_base_patch4_window12_384_22k.pth" --epochs="40" --img_size="480" --plic_pos_temp="0.1" --plic_neg_temp="0.1" --plic_lang_temp="0.3"  --plic_pos_weight="0.5" --plic_neg_weight="0.5" --plic_lang_weight="0.5" --smlm_weight="1.0"  --output-dir="./checkpoints/elia_refcocog_umd"'.format(port) 

print(cmd, flush=True)

os.system(cmd)