I have a question about the size of checkpoint files. Why they all have the same size (519.6 MB). Even I train the program with a tiny version produced checkpoint file has the same volume as you shared it. Is there any specific reason?
The checkpoint size is given by the networks we use, not by dataset size. The number of weights are the same regardless of how many images we use when training them.
I have a question about the size of checkpoint files. Why they all have the same size (519.6 MB). Even I train the program with a tiny version produced checkpoint file has the same volume as you shared it. Is there any specific reason?