TensorFlow Cheat Sheet 2026

The 30 highest-yield TensorFlow facts, distilled from real exam questions. Print it, save it as a PDF, or study it here — free, no sign-up.

1. Which metric should you use when testing a TensorFlow model on a highly imbalanced binary classification dataset? → F1-Score or AUC-ROC
2. Which technique trains a model in multiple steps with increasing dataset sizes? → Curriculum learning
3. What is the purpose of tf.keras.callbacks.CSVLogger during model evaluation training runs? → Appends epoch-level metrics to a CSV file for later analysis
4. What must be true about a model before calling model.load_weights() successfully? → Its layer shapes must match those of the saved weights
5. In TensorFlow, tf.keras.metrics.SparseCategoricalAccuracy is preferred over CategoricalAccuracy when: → Labels are integer class indices rather than one-hot vectors
6. What does dataset.repeat(3) do in tf.data? → Iterates through the dataset 3 times
7. Batch normalization is useful due to? → It normalizes (adjustments) all the input earlier than sending it to the subsequent layer
8. In model.evaluate(), the steps parameter controls what? → Number of batches to draw from the dataset
9. What is mixed precision training in TensorFlow? → Using float16 for computations and float32 for weights to speed up training
10. What does the num_parallel_calls parameter in dataset.map() do? → Parallelizes the map transformation across multiple CPU threads
11. What is the purpose of tf.data.Dataset.batch() when preparing test data? → Groups samples into mini-batches for efficient evaluation
12. What is the primary role of tf.train.CheckpointManager? → Limiting the number of checkpoints kept with max_to_keep
13. Which TensorFlow constant creates a 3×3 identity matrix? → tf.eye(3)
14. What is the purpose of tf.debugging.assert_shapes()? → Verifies tensor shapes match expectations and raises an error if not
15. Which of the following functions should not be used to classify an image at the output layer? → ReLU
16. What is the purpose of the ModelCheckpoint callback? → Saves the model at specified intervals or when performance improves
17. Which function creates a tf.data.Dataset from a list of NumPy arrays? → tf.data.Dataset.from_tensor_slices()
18. Which attribute gives you the number of elements in a TensorFlow tensor? → tf.size()
19. What does model.evaluate() return when a model is compiled with multiple metrics? → A list where the first element is loss followed by each metric value
20. Which TensorFlow Model Analysis (TFMA) concept allows you to evaluate metrics on slices of data? → SlicingSpec
21. What is the purpose of a representative dataset in TFLite quantization? → Calibrates quantization scales for activations to minimize accuracy loss
22. Which TensorFlow function creates a tensor filled with zeros? → tf.zeros()
23. Which low-level TensorFlow function explicitly saves a model in SavedModel format? → tf.saved_model.save()
24. What does the drop_remainder=True argument do in dataset.batch()? → Drops the last batch if it has fewer elements than the batch size
25. Which Keras API allows building models with multiple inputs and outputs? → Functional API
26. Which Keras layer performs 2D convolution for image processing? → Conv2D
27. Which optimizer is recommended for training transformer-based models? → Adam with warmup learning rate schedule
28. What type of object does tf.saved_model.load() return? → A trackable object containing the model's saved functions and variables
29. A ______ is a network that includes feedback linkages from o/p to i/p as well as hidden layers. → Recurrent neural network
30. Which operation performs element-wise multiplication in TensorFlow? → tf.multiply()