t_patch_size issue

[Vincent6896]

Hi. Thank you for your work. When working with the model I encounter a problem that the default value of t_patch_size is 4 and number of frames is 32 in the code, but the value of t_patch_size is 2 in pretrained weight. May I ask how to solve that? Thank you very much.

[vladchimescu]

@Vincent6896 you could simply set t_patch_size = 2 when instantiating your model and loading pretrained weights. For instance, in main_test.py you can provide it as a command line argument (see this line)

In the paper, the authors also used a temporal patch size of 2. I am not sure why it's 4 by default in the code base.