nx_nn_train_gate.nx

nx_nn_train_gate.nx -- proves sovereign fixed-point training (nx_nn_train): batch gradient descent on a linear layer, starting from zero weights, learns a known target mapping -- the MSE loss collapses and the learned weights converge to the target. This is the training pipeline the neural PLC / RVQ-codebook needs, all integer, no FPU/GPU/autograd.