# Run evaluation on the test set for each weight to determine the best weight, only provide the ckpt name from the logs/tmp_sevirlr/checkpoints/...
python -m scripts.train_vae.train_vae_sevirlr --cfg ./scripts/train_vae/cfg.yaml --test --ckpt_name 001.ckpt --gpus 0 1

# Also copy the pretrained vae weight from the github repository to the pretrained_weights/pretrained_weights/vae
python -m scripts.train_vae.train_vae_sevirlr --cfg ./scripts/train_vae/cfg.yaml --test --pretrained --gpus 0 1

# Knowledge Alignment Training
# This step needs a pretrained vae to train the Knowledge Alignment model
# Copy the chosen ckpt from logs/tmp_sevirlr/checkpoints/... (if any) into the directory pretrained_weights/pretrained_weights/vae
# Then modify scripts/train_alignment/cfg.yaml file
# model.vae.pretrained_ckpt_path = <filename of the ckpt file to be used as the pretrained vae, ex: 001.ckpt>
python -m scripts.train_alignment.train_sevirlr_avg_x --cfg ./scripts/train_alignment/cfg.yaml --gpus 0 1