This repository includes tutorials on how to use the TensorFlow estimator APIs to perform various ML tasks, in a systematic and standardised way
APACHE-2.0 License
Please follow the directions in INSTALL if you need help setting up your environment.
Various ML tasks, currently covering:
How to use canned estimators to train ML models.
How to use tf.Transform for preprocessing and feature engineering (TF v1.7)
Implement TensorFlow Model Analysis (TFMA) to assess the quality of the mode (TF v1.7)
How to use tf.Hub text feature column embeddings (TF v1.7)
How to implement custom estimators (model_fn & EstimatorSpec).
A standard metadata-driven approach to build the model feature_column(s) including:
Data input pipelines (input_fn) using:
A standard approach to prepare wide (sparse) and deep (dense) feature_column(s) for Wide and Deep DNN Liner Combined Models
The use of normalizer_fn in numeric_column() to scale the numeric features using pre-computed statistics (for Min-Max or Standard scaling)
The use of weight_column in the canned estimators, as well as in loss function in custom estimators.
Implicit Feature Engineering as part of defining feature_colum(s), including:
How to use the tf.contrib.learn.experiment APIs to train, evaluate, and export models
Howe to use the tf.estimator.train_and_evaluate function (along with trainSpec & evalSpec) train, evaluate, and export models
How to use tf.train.exponential_decay function as a learning rate scheduler
How to serve exported model (export_savedmodel) using csv and json inputs