TensorTonic
Problems

Log Loss (Binary Cross-Entropy)

Compute numerically stable binary log loss by clipping predicted probabilities before cross-entropy evaluation.