LAVT-RIS/donghwa/scripts/submit_test_easyhard.sh

#!/bin/bash

#SBATCH --job-name=lavt_easyhard      # Submit a job named "example"
#SBATCH [email protected]
#SBATCH --mail-type=BEGIN,END,FAIL
#SBATCH --partition=a4000         # a6000 or a100
#SBATCH --gres=gpu:1
#SBATCH --time=7-00:00:00        # d-hh:mm:ss, max time limit
#SBATCH --mem=48000              # cpu memory size
#SBATCH --cpus-per-task=4        # cpu num
#SBATCH --output=log_refcocog_umd_ckpt_testAB.txt           # std output filename

ml cuda/11.0                # 필요한 쿠다 버전 로드
eval "$(conda shell.bash hook)"  # Initialize Conda Environment
conda activate lavt             # Activate your conda environment



# test lavt_one


srun python test.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testA --resume ./checkpoints/ckpt_lavt_one/gref_umd.pth --workers 4 --ddp_trained_weights --window12 --img_size 480
srun python test.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testB --resume ./checkpoints/ckpt_lavt_one/gref_umd.pth --workers 4 --ddp_trained_weights --window12 --img_size 480

# srun python test.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testA --resume ./checkpoints/repro_lavt_one/model_best_gref_umd_lavt_one.pth --workers 4 --ddp_trained_weights --window12 --img_size 480
# srun python test.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testB --resume ./checkpoints/repro_lavt_one/model_best_gref_umd_lavt_one.pth --workers 4 --ddp_trained_weights --window12 --img_size 480

# srun python test_mosaic.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testA --resume ./checkpoints/random_550_lavt_one/model_best_mosaic_gref_umd_lavt_one.pth --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/random_550.yaml
# srun python test_mosaic.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testB --resume ./checkpoints/random_550_lavt_one/model_best_mosaic_gref_umd_lavt_one.pth --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/random_550.yaml

# srun python test_mosaic.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testA --resume experiments/refcocog_umd/random_gref_umd_460_40epoch_2/model_best_random_gref_umd_460_40epoch_2.pth --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/random_460.yaml
# srun python test_mosaic.py --model lavt_one --swin_type base --dataset refcocog --splitBy umd --split testB --resume experiments/refcocog_umd/random_gref_umd_460_40epoch_2/model_best_random_gref_umd_460_40epoch_2.pth --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/random_460.yaml


# retrieval
# srun python test_mosaic_retrieval.py --model lavt_one --swin_type base --dataset refcocog \
#     --splitBy umd --split testA --resume experiments/refcocog_umd/retrieval_filter_gref_umd_433_10up_top200/model_best_retrieval_filter_gref_umd_433_10up_top200.pth  \
#     --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/retrieval_433_10up.yaml

# srun python test_mosaic_retrieval.py --model lavt_one --swin_type base --dataset refcocog \
#     --splitBy umd --split testB --resume experiments/refcocog_umd/retrieval_filter_gref_umd_433_10up_top200/model_best_retrieval_filter_gref_umd_433_10up_top200.pth  \
#     --workers 4 --ddp_trained_weights --window12 --img_size 480 --config config/retrieval_433_10up.yaml