GPU vs CPU tensor placement in TensorFlow - Quick Revision & Key Differences

Recall & Review

beginner

What is the main difference between CPU and GPU in tensor placement?

CPU handles general tasks and is good for sequential operations, while GPU is specialized for parallel tasks, making it faster for large tensor operations in machine learning.

beginner

How does TensorFlow decide where to place a tensor by default?

TensorFlow tries to place tensors on the GPU if available for faster computation; otherwise, it uses the CPU.

intermediate

What is the TensorFlow command to explicitly place a tensor on the GPU?

Use with tf.device('/GPU:0'): to place tensors or operations explicitly on the first GPU device.

intermediate

Why might you want to place tensors on the CPU instead of the GPU?

You might place tensors on the CPU if the GPU memory is limited, or for operations that are not efficient on GPU, like small or simple computations.

advanced

What happens if you try to perform an operation on tensors placed on different devices?

TensorFlow will automatically copy data between devices, but this can slow down performance due to data transfer overhead.

Which device is generally faster for large matrix multiplications in TensorFlow?

A Hard disk

B GPU

C CPU

D RAM

GPUs are designed for parallel processing, making them faster for large matrix operations.

How do you specify a tensor to be placed on the CPU in TensorFlow?

A with tf.device('/DISK:0'):

B with tf.device('/GPU:0'):

C with tf.device('/TPU:0'):

D with tf.device('/CPU:0'):

The '/CPU:0' device string places tensors on the CPU.

What is a downside of placing tensors on different devices for operations?

A Tensors get deleted

B TensorFlow crashes

C Automatic data transfer slows performance

D No downside

Data transfer between devices adds overhead and slows down computation.

If no GPU is available, where does TensorFlow place tensors by default?

A CPU

B GPU

C TPU

D Disk

TensorFlow defaults to CPU if no GPU is found.

Which TensorFlow function helps you check available devices?

A tf.config.list_physical_devices()

B tf.device()

C tf.Tensor()

D tf.run()

tf.config.list_physical_devices() lists available hardware devices like CPU and GPU.

Explain how TensorFlow manages tensor placement between CPU and GPU and why this matters.

Think about speed and where computations happen.

Describe a situation where placing tensors on CPU might be better than GPU.

Consider resource limits and operation size.