The running process:
0. download the cifar dataset and put in ./data/
1. Ensemble-in-One training: python train.py --warmup (wramup is optional) 
2. Path extraction: python derive_one_path.py (modify the path of EIO model), then you can find the derived model in 'tmp_file'
3. Fine-tune: python finetune.py, then you can get the finetuned model in 'save_temp'
4. Pack the model for evaluation: python pack_model_for_test.py
5. move the model.pth to ensemble_in_one/checkpoint/eio/seed_0/ResNet_eps_0.07, you can also change a dir
6. Run test in EIO_test: cd EIO_test/script/, sh evaluation.sh

