Zeta36 / chess-alpha-zero

Chess reinforcement learning by AlphaGo Zero methods.
MIT License
2.13k stars 481 forks source link

Update BN axis #90

Closed CamilleChiquet closed 4 years ago

CamilleChiquet commented 4 years ago

If channels first, then batch normalization should be on the axis 1 (channel axis). "axis=-1" is in the case where channels are on the last dimension (batch, width, height, channels).