Skip to main content

Estimator API

Estimator API - Wrapping a model in a common set of APIs
Create production ready machine learning models
Train on large datasets that doesn't fit in memory
Monitor your training metrics in tensorboard

Dataset API

All the tf. commands that we write do not actually process any data, they just build graphs

map is used for one to one transformation (parsing a single line into features and label)
flat_map is used for one to many transformation (to flatten all the resulting text lines into one dataset

Data Parallelism = replicate your models on multiple workers

Monitoring with TensorBoard

Serving an ML model

Scaling TensorFlow models with CMLE (Cloud Machine Learning Engine)

http://onlinetutorials.today/tensorflow/tensorflow-interview-questions-and-answers

Dataset API
All the tf. commands that we write do not actually process any data, they just build graphs
Data Parallelism = replicate your models on multiple workers
Monitoring with TensorBoard
Serving an ML model
Scaling TensorFlow models with CMLE (Cloud Machine Learning Engine)