Machine Learning Models
The Analytics module allows you to execute your Machine Learning (ML) models “run time”. You can feed live data to the model and get live output from it.
Note: To work with ML models in Analytics, your computer has to meet certain requirements. If it does not, the ML-specific processors will not be available for flow building.
To upload a model to Analytics, you need a SavedModel format from TensorFlow. Once you have created and saved a model, you must compress and zip it prior to uploading it to Analytics. SavedModel is a directory containing serialized signatures and the state needed to run them, including variable values and vocabularies. The SavedModel (saved_model.pb) file stores the TensorFlow model and a set of named signatures, each identifying a function that accepts tensor inputs and produces tensor outputs. SavedModels may contain multiple variants of the model.
Once your ML model is ready, proceed as follows:
- Create an analytics flow where the function processor is based on your model.
- TensorFlow Images processor for images. See Create an Analytics Flow and TensorFlow Images Processor for details.
- (Optional) Add the Tengo Script processor to modify the model output. See Tengo Script for details.
To access Analytics Models:
- Log in to Manufacturing Connect Edge.
From the Navigation panel, navigate to Analytics > Models. The Models pane appears.
Analytics Models pane