(Day 34) Deploying ML models and TensorBoard

Ivan Ivanov · February 4, 2024

Hello :) Today is Day 34!

A quick summary of today:

When creating an app for iOS/Android, it would be nice to follow the following step-by-step guide.

image

Set up the model in the desired platform > process input provided by the user > pass the input to the model > get results and show them

Then, I learned about tensorflow datasets (tfds).

image

Cache can be used to improve performance

image

Memory cache provides faster access time and fits well with small to medium-sized datasets. Disk cache provides persistence and scalability and fits well with large datasets

Next, I learned a bit about TensorBoard

image

You can see the mean and standard deviation for each epoch on the TensorBoard (Cool!)

image

In addition, if you create a classification model, you can also see the confluence matrix of each epoch

For example, epoch 1:

image

Epoch 3:

image

Lastly, learned about Federated learning

Federated learning is a technology that learns a global model in a decentralized situation in which a central server and local clients cooperate with each other. It has advantages in terms of data privacy and communication efficiency, and it requires solutions to problems such as completely decentralized learning, and there are several challenges such as model learning in situations where data is unevenly distributed.

image

image

Actually, I was going to certify TF developer today, but my card kept getting rejected :/ (I emailed Google)

If I cannot do that, my next options are:

I have to decide tomorrow


That is all for today!

See you tomorrow :)

Original post in Korean