TensorFlow Developer Practice Test

Answer choices

1. A. Flatten
2. B. Dense
3. C. MaxPooling2D
4. D. BatchNormalization

Wrong-answer review

• A. Flatten: Flatten collapses dimensions rather than pooling local spatial windows.
• B. Dense: Dense layers operate on vectors and do not perform spatial pooling.
• D. BatchNormalization: BatchNormalization normalizes activations and does not pool spatial features.

Answer choices

1. A. Create dataset of files -> prefetch -> batch -> map -> flat_map
2. B. Create dataset of files -> map -> batch -> flat_map -> prefetch with static buffer size 1
3. C. Create dataset of files -> batch -> prefetch -> map -> flat_map
4. D. Create dataset of files -> flat_map with parallel reads -> map with num_parallel_calls -> batch -> prefetch(buffer_size=tf.data.AUTOTUNE)

Answer choices

1. A. An open-source machine learning platform for building, training, and deploying models
2. B. A spreadsheet tool used only for manual budgeting
3. C. A network firewall product with no model-development features
4. D. A database migration utility that cannot run ML workloads

Answer choices

1. A. A probability-like score for the positive class
2. B. An image segmentation mask
3. C. A sequence of embedding vectors
4. D. A time-series window index

Answer choices

1. A. Data leakage
2. B. Gradient clipping
3. C. Model checkpointing
4. D. Image augmentation

Answer choices

1. A. The TensorFlow Certificate exam has been closed while TensorFlow evaluates the next step in the certificate program
2. B. The old exam is open continuously with no changes to the program
3. C. The exam has been replaced by a required live interview with the TensorFlow core team
4. D. The certificate page says nothing about the exam's current availability

Answer choices

1. A. .keras
2. B. .css
3. C. .dns
4. D. .pptx

Answer choices

1. A. Increase the notebook font size
2. B. Rename the model file after each epoch
3. C. Replace every medical record with an image
4. D. Add metrics that better reflect minority-class performance, such as precision and recall

Answer choices

1. A. Embeddings automatically remove the need for labels
2. B. Embeddings convert text into image filters
3. C. Embeddings guarantee perfect grammar in generated text
4. D. Embeddings learn dense representations that capture relationships between tokens

Answer choices

1. A. It prevents the model from loading pretrained ImageNet weights
2. B. It removes all feature-extraction layers and keeps the original classifier
3. C. It disables all batch normalization layers during fine-tuning
4. D. It excludes the final dense classifier so a custom head can be added