TensorFlow Processor
The TensorFlow processor feeds time-series data to a TensorFlow model you've already created and trained.
Why use the TensorFlow processor
Use this processor when your analytics flow needs to pass time-series values, rather than image data, into a TensorFlow model, for example, running an existing predictive-maintenance model against a stream of sensor readings.
How it works
Point Model Path at your trained TensorFlow model. Tags, Input Operation, and Output Operation must match the names used when the model was created: Tags identifies which model to load if there's more than one, and Input Operation and Output Operation identify the tensors the model expects as input and returns as output.
Number of Inputs sets how many inputs are passed to the model. Window Size sets the size of the values array to process. Time Shift Ms shifts the output's timestamp by the specified number of milliseconds.
Parameters
Parameter | Details |
|---|---|
Model Path | The file path of the pre-trained TensorFlow model. |
Number of Inputs | The number of inputs given to the TensorFlow model. |
Tags | Tags for loading the model. Separate multiple tags with commas. |
Input Operation | The input tensor used when the model was created, usually the dictionary of values fed to the model. |
Output Operation | The output tensor used when the model was created, usually the model's output. |
Window Size | The size of the values array to process. |
Time Shift Ms | How many milliseconds to shift the current timestamp by. |
Pass Through Value | Determines whether the processor's output is the model's result or the original unchanged input value. |