No need of optimization across neural network library and mcts part
This can be seen as an optimization boundary, like what we did for client cards
This wrapper cpp can also deal with header workarounds for tiny dnn library
We can also define callback interface to get board values
This callback should be designed by template concept
No need of optimization across neural network library and mcts part This can be seen as an optimization boundary, like what we did for client cards This wrapper cpp can also deal with header workarounds for tiny dnn library
We can also define callback interface to get board values This callback should be designed by template concept