cnaseeb/aihub

aihub's Issues

Tensorflow versions expect different parameter names

When providing the argument for evaluating the threshold to stop the training, we provide accuracy
and when compiling the model using evaluation metrics accuracy we use 'accuracy' but this depends on which version of the tensorflow library you are using. If you are using Tensorflow version (2.1.0 or above), you should use the full name of the parameter ('accuracy') otherwise you will get a strange error which points to the threshold but has nothing to do with it. Similarly, if you are using lower versions (1.13 etc), use the shorthand 'acc' and your model will work correctly.

It appears to be the fact that grader is using older version of tensorflow that is why you have to use acc instead of accuracy to pass the grader evaluation of your submission.