Run the following command to train Earthformer on SEVIR dataset. Change the configurations in cfg_sevir.yaml
MASTER_ADDR=localhost MASTER_PORT=10001 python train_cuboid_sevir.py --gpus 2 --cfg cfg_sevir.yaml --ckpt_name last.ckpt --save tmp_sevir
Or run the following command to directly load pretrained checkpoint for test.
MASTER_ADDR=localhost MASTER_PORT=10001 python train_cuboid_sevir.py --gpus 2 --pretrained --save tmp_sevir
Run the tensorboard command to upload experiment records
tensorboard dev upload --logdir ./experiments/tmp_sevir/lightning_logs --name 'tmp_sevir'
Run the following command to train Earthformer on SEVIR-LR dataset. Change the configurations in cfg_sevirlr.yaml
MASTER_ADDR=localhost MASTER_PORT=10001 python train_cuboid_sevir.py --gpus 2 --cfg cfg_sevirlr.yaml --ckpt_name last.ckpt --save tmp_sevirlr
Run the tensorboard command to upload experiment records
tensorboard dev upload --logdir ./experiments/tmp_sevirlr/lightning_logs --name 'tmp_sevirlr'