In Floyd book it says that 2 input bias currents for 2 inputs ideally equal. I understand if they are equal in a common mode, but how they can be equal in a differential mode if in this case, one input is connected to the ground, or voltages are in opposite phase which at least gives opposite directions of the currents?
The basic error you made is in your understanding of bias currents: -
Bias currents are not something that YOUR input signal is responsible for so, the direction or polarity of input signals are irrelevant. Bias currents can be regarded as currents taken (or given) by the input stage of the OP-AMP and they will flow into or out of your input voltage sources irrespective of their polarity. They are self-generated within the op-amp and not to be seen as some current proportional to your input signal level.