A question about inference speed.

[Baek-sohyeon]

I am a student studying Sage Maker Neo.
I am working on this tutorial.

• Training and Serving with TensorFlow on Amazon SageMaker
• https://github.com/aws/amazon-sagemaker-examples/blob/master/aws_sagemaker_studio/frameworks/tensorflow_mnist/tensorflow_mnist.ipynb

What I'm curious about is that the inference speed is similar when using the c5 instance and when using the p2 instance.
Please let me know what I am missing.