Residual-networks Wrong values when training=True

[tabassum]

This is my code:

This is the error:

where am i making the mistake? looks fairly straight forward

[TMosh]

Try specifying the kernel size in all of the Conv2D() layers.

[TMosh]

Your shortcut path is missing the “training=training” argument in batch normalization.

[tabassum]

that was ignorant of me, thank you. did the trick