NVIDIA Morpheus (24.06)
(Latest Version)

DFP Rolling Window Module

This module is responsible for maintaining a rolling window of historical data, acting as a streaming caching and batching system.




Example Value

Default Value

cache_mode string Mode for managing user cache. Setting to batch flushes cache once trigger conditions are met. Otherwise, continue to aggregate user’s history. “batch” “batch”
trigger_on_min_history int Minimum history to trigger a new training event 1 1
trigger_on_min_increment int Minmum increment from the last trained to new training event 0 0
timestamp_column_name string Name of the column containing timestamps “timestamp” “timestamp”
aggregation_span string Lookback timespan for training data in a new training event “60d” “60d”
cache_to_disk bool Whether or not to cache streaming data to disk false false
cache_dir string Directory to use for caching streaming data “./.cache” “./.cache”

{ "cache_mode": "batch", "trigger_on_min_history": 1, "trigger_on_min_increment": 0, "timestamp_column_name": "timestamp", "aggregation_span": "60d", "cache_to_disk": false, "cache_dir": "./.cache" }

Previous dfp_preproc
Next DFP Split Users Module
© Copyright 2024, NVIDIA. Last updated on Jul 8, 2024.