#!/bin/bash # sh scripts/ood/csi/cifar10_test_ood_csi.sh GPU=1 CPU=1 node=36 jobname=openood PYTHONPATH='.':$PYTHONPATH \ # srun -p dsta --mpi=pmi2 --gres=gpu:${GPU} -n1 \ # --cpus-per-task=${CPU} --ntasks-per-node=${GPU} \ # --kill-on-bad-exit=1 --job-name=${jobname} -w SG-IDC1-10-51-2-${node} \ python main.py \ --config configs/datasets/cifar10/cifar10.yml \ configs/datasets/cifar10/cifar10_ood.yml \ configs/networks/csi_net.yml \ configs/pipelines/test/test_ood.yml \ configs/postprocessors/msp.yml \ configs/preprocessors/base_preprocessor.yml \ --network.pretrained True \ --network.checkpoint 'results/cifar10_csi_net_csi_step2_e100_lr0.1/s0/best.ckpt' \ --merge_option merge ############################################ # alternatively, we recommend using the # new unified, easy-to-use evaluator with # the example script scripts/eval_ood.py # especially if you want to get results from # multiple runs python scripts/eval_ood.py \ --id-data cifar10 \ --root ./results/cifar10_csi_net_csi_step2_e100_lr0.1 \ --postprocessor msp \ --save-score --save-csv