Benchmarking Machine Learning Models for Quantum Error Correction