ntakouris/tfx-pipelines-playground

End to End Machine Learning Pipelines using Tensorflow Extended.

TFX Pipelines Playground

This is basically the TFX Example of chicago taxi trips pipeline with some extra stuff going on like Tuning and some notebooks for visualization.

Repository Contents

The pipeline that is implemented is the: TFX Chicago Taxi Pipeline Example, slightly modified.

File	Contents
data	The CSV raw data for the taxi pipeline example
pipeline.py	The full, vanilla TFX pipeline declaration and component configuration
model.py	Preprocessing, Model Creation, Training, Hyperparameter tuning
example_data_visualization.ipynb	ExampleValidator and StatisticsGen: Raw data and schema visualization notebook through TFDV
model_analysis_visualization.ipynb	Evaluator TFMA result visualization notebook
tensorboard_training_visualization.ipynb	Trainer training logs visualization notebook
tf_serving_communication.ipynb	Parse a feature dict to tf.Example and perform a POST request to obtain predictions (notebook)
mlmd_artifacts_from_pipeline_info.ipynb	Query the MLMD store by pipeline info to get all artifacts by type
launch_model_local.sh	Utility for quick launch of a tf serving docker container