Some errors happened in the evaluate mode in test.py

[BugMaker2002]

When I try to run the following command for evaluate mode in test.py:
python test.py --evaluate --gpu 0 --epochs 100 --batch_size 128 --lr 5e-3 --dropout 0
--pretrained "./logs/ubfc/train/ubfc_best.pth"
--dataset_name "ubfc-rppg"
--dataset_dir "/share2/data/zhouwenqing/UBFC_rPPG_processed"
--workers 4 --vid_frame 150 --vid_frame_stride 2
--log_dir "./logs/ubfc/test"
--model_depth 18
Some errors happened:

However, if I remove the evaluate parameter from the command like this: the program runs without error
python test.py --gpu 0 --epochs 100 --batch_size 128 --lr 5e-3 --dropout 0
--pretrained "./logs/ubfc/train/ubfc_best.pth"
--dataset_name "ubfc-rppg"
--dataset_dir "/share2/data/zhouwenqing/UBFC_rPPG_processed"
--workers 4 --vid_frame 150 --vid_frame_stride 2
--log_dir "./logs/ubfc/test"
--model_depth 18

[Dylan-H-Wang]

Because the provided weights are SSL-pre-trained weights and it is expected to be fine-tuned on some datasets.
The '--evaluate' flag is used for inference which means the weights for--pretrained should be the result of fine-tuned model.
In a word, run test.py with provided pre-trained weights without --evaluate flag first. Then, you can run test.py with '--evaluate' using the output weights from the last test.py run.