Note: This is documentation for version 5.12 of Source. For a different version of Source, select the relevant space by using the Spaces menu in the toolbar above

Insight Replicate Runner

Owned by Kuan Bartel (Deactivated)
Last updated: 12 Oct 2023 by TeamcityVersion comment

Overview

The Insight Replicate Runner is a combination of the Insight Runner and the Replicate Runner. It performs an Insight run where each population individual is run multiple times using different timeseries inputs generated from the Time Series Cycle Creator (TSCC). The runner can be used, for example, to optimise an eWater Source model with a range of alternative climate data inputs and decision variables.

Note that, it is possible to achieve similar behaviour using the Insight Runner and the Load Time Series option of Insight to load a set of pre-generated timeseries files.

eWater Source and Insight Project Setup

To run an Insight Replicate job, the eWater Source project must already be created with the timeseries to be cycled loaded in Data Sources (see Replicate Runner for details). An Insight configuration file called config.txt must also be included (see Configuring an Insight Optimisation Run).

The Load Time Series option of Insight is not supported by the Insight Replicate Runner as the setting will be overwritten by the generated, cycled timeseries data.

Once configured, the eWater Source project(s), the input timeseries csv files (the files to be cycled), and the Insight configuration file must be placed in a zip archive file. The Insight configuration file must be renamed "config.txt" if it is saved using a different name in Insight.

Note that the Insight config file must be in the root of the zip archive and not within a folder. The project setting in the config file must refer to the folder which contains the config file otherwise Source project and timeseries files may not be located correctly and cause the run to fail.

Example Insight Replicate project zip file

InsightReplicateExample.zip

In this example there is a single project, Multi_obj_optimise.rsproj, and there is one input timeseries file, Inflows.csv.

  • InsightReplicateExample.zip
    • config.txt
    • Inflows.csv
    • Multi_obj_optimise.rsproj

Insight config.txt paths:

project C:\InsightReplicateExample

Note the 'timeseries' parameter is not specified. It will automatically be specified by the Insight Replicate Runner, replacing any existing value.

The Run Manager Cloud Application is currently configured to use multiple versions of Source and has two plugins available:

  • UrbanDeveloper
  • SubSource

Before running projects on Run Manager it is always recommended to save the projects using a supported version to avoid needing to upgrade the projects during Source runs.

Contents

Job Creation

Upload Data

Click "Choose Insight zip" and select the .zip file that contains your Insight Replicate project. Depending on the size of the file it may take a moment for the file to upload.

Once the project has finished being uploaded, the Insight Replicate Runner configuration options will be displayed.

Configuration

The configuration on the Insight Replicate Runner combines the same options as both the Insight Runner and the Replicate Runner. Please refer to those runners for details about configuration.

Queue 

Table 1 lists the queue options for the Insight Replicate Runner.

Table 1. Queue options.

Configuration Option

Description

Job Name

The name given to the job when viewed in the Run Manager

Agent End pointsChoose which type of agents should be used to run the job (will be hidden if Run Manager is configured to only have one type of agent endpoint)
Source VersionChoose the Source version to use for the Job
Number of AgentsThe number of agents to be allocated to the job. Note that Insight Jobs must have at least 2 agents (one master and one endpoint). You may specify any number of agents however to minimise the waiting time of agents the number of agents should be a factor of the runs per generation plus one for the Insight master, where runs per generation = (population size × project count × number of replicates), as NSGA-II cannot move onto the next generation until all individuals in the current generation have finished.
Model Run Estimate (Minutes)(Optional) The user's estimate of how long it will take to run a single individual = (project count × number of replicates)
Estimate(Optional) Estimates the costs of running the job based on the user's Model Run Estimate, the number of agents, the population size, and the number of generations.
QueueAdds the job to the queue of jobs to be run

Results

The Insight Replicate Runner will provide the following result files in a zip:

  • job.log - Log file containing information and messages generated during the Job
  • results.csv - The Insight results file which can be imported into Insight to be viewed
  • cache.csv - The Insight results cache file which can be imported into Insight to be viewed and used in subsequent runs
  • A folder for each replicate containing replicate input timeseries files used for the run in the same folder structure as provided in the original zip file