Progress tracking and reporting in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Metrics

intermediate

Understanding Training Loss Reporting

During training a neural network, the loss value reported after each epoch is 0.05. What does this number represent?

A The average error between the model's predictions and true values over the training data for that epoch.

B The percentage of correct predictions made by the model on the training data.

C The total number of training samples processed in that epoch.

D The time in seconds it took to complete the epoch.

💡 Hint

Loss measures how far off the model's guesses are from the real answers.

Loss is a number that tells us how wrong the model's predictions are on average. A lower loss means better predictions.

❓ Predict Output

intermediate

Output of Progress Reporting Code

What will be printed by this Python code snippet that tracks training progress?

Agentic AI

epochs = 3
for epoch in range(1, epochs + 1):
    loss = 0.1 / epoch
    print(f"Epoch {epoch}: Loss = {loss:.3f}")

A

Epoch 1: Loss = 1.000
Epoch 2: Loss = 0.500
Epoch 3: Loss = 0.333

B

Epoch 1: Loss = 0.100
Epoch 2: Loss = 0.200
Epoch 3: Loss = 0.300

C

Epoch 1: Loss = 0.100
Epoch 2: Loss = 0.050
Epoch 3: Loss = 0.033

D

Epoch 1: Loss = 0.010
Epoch 2: Loss = 0.005
Epoch 3: Loss = 0.003

💡 Hint

Loss decreases as epoch number increases because it is divided by epoch.

The loss is calculated as 0.1 divided by the epoch number, so it decreases each epoch.

❓ Model Choice

advanced

Choosing a Model for Progress Tracking

You want to build a system that reports training progress with detailed metrics like accuracy, loss, and validation scores after each epoch. Which model type is best suited for this?

A A simple linear regression model without callbacks or metric tracking.

B A deep learning model using a framework that supports callbacks to report metrics after each epoch.

C A clustering model that groups data points without training epochs.

D A rule-based system that does not learn from data.

💡 Hint

Progress tracking needs a model that trains in steps and reports metrics.

Deep learning frameworks often provide callbacks to report metrics like loss and accuracy after each epoch, enabling detailed progress tracking.

❓ Hyperparameter

advanced

Impact of Batch Size on Progress Reporting

How does increasing the batch size during training affect the frequency and granularity of progress reports?

A Larger batch size means fewer updates per epoch, so progress reports are less frequent and less granular.

B Larger batch size causes progress reports to be skipped entirely.

C Batch size does not affect progress reporting frequency or granularity.

D Larger batch size means more frequent progress updates with finer granularity.

💡 Hint

Batch size controls how many samples are processed before updating the model.

With larger batches, fewer updates happen per epoch, so progress reports based on batch updates occur less often.

🔧 Debug

expert

Debugging Incorrect Progress Metrics

A training script prints validation accuracy after each epoch, but the values stay constant at 0.5 and never improve. What is the most likely cause?

Agentic AI

for epoch in range(5):
    train_model()
    val_acc = evaluate(validation_data)
    print(f"Epoch {epoch+1} Validation Accuracy: {val_acc}")

A The model is not learning because training data is not shuffled each epoch.

B The evaluate function is called before training updates the model weights.

C The validation dataset is empty, so accuracy defaults to 0.5.

D The evaluate function always returns a fixed value of 0.5 due to a bug.

💡 Hint

Check if the evaluation function returns dynamic results or a constant.

If validation accuracy never changes, the evaluation function might be returning a constant value due to a bug.