GRU layer in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output shape of a GRU layer

What is the output shape of the following GRU layer when input shape is (batch_size=32, timesteps=10, features=8)?

TensorFlow

import tensorflow as tf
layer = tf.keras.layers.GRU(16)
input_shape = (32, 10, 8)
import numpy as np
x = np.random.random(input_shape).astype('float32')
out = layer(x)
print(out.shape)

A (32, 10, 16)

B (32, 16)

C (10, 16)

D (32, 8)

💡 Hint

By default, GRU returns the last output for each sample, not the full sequence.

The GRU layer with default settings returns the last output for each sample, so the output shape is (batch_size, units) which is (32, 16).

❓ Model Choice

intermediate

Choosing GRU for sequence data

You want to build a model to predict the next word in a sentence using a recurrent neural network. Which layer is best suited for capturing long-term dependencies efficiently?

A SimpleRNN layer

B Conv2D layer

C Dense layer

D GRU layer

💡 Hint

GRU is designed to handle long-term dependencies better than SimpleRNN.

GRU layers are designed to capture long-term dependencies in sequences more efficiently than SimpleRNN layers, making them suitable for language modeling.

❓ Hyperparameter

advanced

Effect of return_sequences in GRU

What is the effect of setting return_sequences=True in a GRU layer?

A The GRU disables the reset gate.

B The GRU returns only the last output, producing a 2D tensor.

C The GRU returns output for each timestep, producing a 3D tensor.

D The GRU output shape becomes (batch_size, features).

💡 Hint

Think about whether the output includes all timesteps or just the last one.

Setting return_sequences=True makes the GRU return the output at every timestep, resulting in a 3D tensor of shape (batch_size, timesteps, units).

🔧 Debug

advanced

Identifying error in GRU input shape

What error will this code raise and why? import tensorflow as tf layer = tf.keras.layers.GRU(10) x = tf.random.uniform((5, 10)) out = layer(x)

TensorFlow

import tensorflow as tf
layer = tf.keras.layers.GRU(10)
x = tf.random.uniform((5, 10))
out = layer(x)

A ValueError: Input 0 of layer 'gru' is incompatible with the layer: expected 3D tensor, got 2D tensor instead.

B RuntimeError: GRU layer cannot be called without batch size.

C No error, outputs shape (5, 10).

D TypeError: GRU layer requires integer units argument.

💡 Hint

Check the input tensor dimensions expected by GRU layers.

GRU layers expect 3D input tensors (batch_size, timesteps, features). The input here is 2D, causing a ValueError.

❓ Metrics

expert

Interpreting GRU training loss and accuracy

A GRU model for binary classification is trained for 10 epochs. The training loss steadily decreases, but training accuracy remains around 50%. What is the most likely explanation?

A The model is not learning; output layer or loss function might be incorrect.

B The model is overfitting the training data.

C The learning rate is too high causing unstable training.

D The dataset is perfectly balanced and the model is performing well.

💡 Hint

If loss decreases but accuracy stays at chance level, what could be wrong?

If loss decreases but accuracy stays near 50%, the model might be outputting probabilities but the final classification or loss function is misconfigured.