You have developed a scikit-learn classifier model and now want to migrate it to TensorFlow. The plan is to use the same training dataset used for the scikit-learn model to train the new TensorFlow classifier model. After training, you aim to compare the performances of both models using a common test set. You decide to use the Vertex AI Python SDK to manually log evaluation metrics, specifically focusing on the F1 scores and confusion matrices of each model. What is the correct way to log these metrics using the Vertex AI Python SDK?