How to perform validation of domain generalization？

[freya-7]

I have trained the model on imagenet and want to validate it on the other four variants. But I didn't find the relevant xd_test.bash file, could you provide it?

[htyao89]

I have trained the model on imagenet and want to validate it on the other four variants. But I didn't find the relevant xd_test.bash file, could you provide it?

The saved model's path might be changed with your path (--model-dir).
"--load-epoch" is set as the epoch of training.

#!/bin/bash

cd ..

custom config

DATA=/data/yht/data/cl/data/
TRAINER=KgCoOp

CFG=vit_b16_ep100_ctxv1
SHOTS=16
WEIGHT=8.0

for DATASET in imagenetv2 imagenet_sketch imagenet_a imagenet_r
do
for SEED in 1 2 3
d
DIR=output_0122_cd/evaluation/${TRAINER}/${CFG}${SHOTS}shots/${DATASET}/seed${SEED}
if [ -d "$DIR" ]; then
echo "Oops! The results exist at ${DIR} (so skip this job)"
else
python train.py
--root ${DATA}
--seed ${SEED}
--trainer ${TRAINER}
--dataset-config-file configs/datasets/${DATASET}.yaml
--config-file configs/trainers/${TRAINER}/${CFG}.yaml
--output-dir ${DIR}
--model-dir output_0122_cd/base2new/train_base/imagenet/shots${SHOTS}_${WEIGHT}/${TRAINER}/${CFG}/seed${SEED}
--load-epoch 25
--eval-only
TRAINER.COOP.N_CTX 4
fi
done
done