hunkim
Save directory checking?
Should we put the save directory under the data directory? Or at least, we need to check if the directory exists.
Ckpt.model_checkpoint_path: indicates the location where the model is stored, does not need to provide the name of the model, because it will check the checkpoint file to see which one is the latest and what does it called.
In source code saver.py the method get_checkpoint_state will check that the file exists before opening it. so I think there is no need to put the save directory under the data directory.
So you want to put the results in the input data folder? Could you make a pull request?
The result set for each training session is automatically saved in the save / model path, and is not overwritten as long as the new model name is not reassigned. So there is no need to put the save file under the data directory.
If we have two input data directories and run training twice, I think it will overwrite, right? On Fri, 4 Nov 2016 at 3:43 PM, sunxiaobiu [email protected] wrote:
The result set for each training session is automatically saved in the save / model path, and is not overwritten as long as the new model name is not reassigned. So there is no need to put the save file under the data directory.
— You are receiving this because you were assigned.
Reply to this email directly, view it on GitHub https://github.com/hunkim/word-rnn-tensorflow/issues/18#issuecomment-258363472, or mute the thread https://github.com/notifications/unsubscribe-auth/AA3DV-4bnC9HJPX6Vl6VxOWyzFdP7rc6ks5q6uITgaJpZM4KX_Vl .