MLA-C01 Question 169
Select 3You are training a deep learning model on AWS SageMaker for a text classification task. While reviewing the model's performance, you notice the model size is too large, causing deployment latency issues. Which of the following factors directly influence the size of your model?
- A
The number of layers in the neural network
- B
The size of the training dataset
- C
The number of parameters in the model
- D
The precision used for representing weights (e.g., FP32 vs FP16)
- E
The type of optimizer used during training
Show answer and explanation
Correct answers: A, C, D
Explanation
The size of a machine learning model is influenced by the architecture (e.g., number of layers and parameters), and how weights are represented (e.g., numerical precision). While other factors like the training dataset size and optimizer type are important for training quality and efficiency, they do not directly determine the model's storage size or memory usage.
- A. Correct.
The number of layers in the neural network impacts model size because each layer adds additional parameters, increasing the memory required to store the model.
- B. Incorrect.
The size of the training dataset does not directly affect the model size. It influences the training process but not the storage size of the model itself.
- C. Correct.
The number of parameters in the model directly determines the size of the model since parameters (weights and biases) occupy memory.
- D. Correct.
The precision used for representing weights, such as FP32 (32-bit floating point) vs FP16 (16-bit floating point), directly affects the memory footprint of the model. Lower precision reduces model size.
- E. Incorrect.
The type of optimizer used during training influences the learning process but does not directly impact the size of the final model.