Tutorial: using NeurEco command line interface for a Tabular Regression problem#

NeurEcoDNN is the executable used for building, evaluating and exporting Tabular models (Regression, Classification, Compression). The executable can be called directly from a terminal / powershell only after a full installation.

Note

When using a portable version of the software, make sure to add its bin directory to the environment variable PATH.

To call the executable, simply run the following command:

neurecoDNN

which will output:

     _   __                ______
    / | / /__  __  _______/ ____/________
   /  |/ / _ \/ / / / ___/ __/ / ___/ __ \
  / /|  /  __/ /_/ / /  / /___/ /__/ /_/ /
 /_/ |_/\___/\__,_/_/  /_____/\___/\____/
                 === A D A G O S ===

Version: 4.01.2591.0 Compiled with MSVC v1928  Dec  5 2022 Matlab runtime:no
OpenMP: yes
MKL: yes
Version Ref: 27284d298a51ac68c0443ce3e5caee63cd26acb0
usage: neurecoDNN [-h] [command <parameters>]

Entry point for NeurEco model building and evaluation.

Commands:
build <configurationFilename>
          build a neureco model from a given input solution/input set.
evaluate <configurationFilename>
          evaluate a deepROM model from a given excitation.
exportC <NeurecoFilename path> <CFilename path> <precision>
exports NeurecoFilename model as an .h file

exportFMU <NeurecoFilename path> <fmuFilename path> <platform identifier>
    export NeurecoFilename model as an FMU file
  platform: 1=windows, 2=linux, 3=both, default: both.

exportONNX <NeurecoFilename path> <ONNXFilename path> <precision>
exports NeurecoFilename model as an ONNX file

exportVBA <NeurecoFilename path> <VBA Filename path> <precision>
exports NeurecoFilename model as an .bas file

...

Optional arguments:
-h, --help   show this message and exit

The following section uses the test case Energy consumption. This test case is delivered with the NeurEco installation package.

To build a Tabular Regression model using the executable:

{
  "neurecoDNN_build": {
    "DevSettings": {
      "disconnect_inputs_if_possible": true,
      "final_learning": true,
      "initial_beta_reg": 0.1,
      "parameter_number_limit": 0,
      "valid_percentage": 33.33
    },
    "UserSettings": {
      "gpu_id": 0,
      "use_gpu": false
    },
    "build_compress": false,
    "checkpoint_address": "./EnergyConsumption/EnergyConsumption.checkpoint",
    "classification": false,
    "exc_filenames": [
      "x_train.csv"
    ],
    "freeze_structure": false,
    "input_normalization": {
      "normalize_per_feature": true,
      "scale_type": "max_centered",
      "shift_type": "min_centered"
    },
    "output_filenames": [
      "y_train.csv"
    ],
    "output_normalization": {
      "normalize_per_feature": true,
      "scale_type": "auto",
      "shift_type": "auto"
    },
    "resume": false,
    "starting_from_checkpoint_address": "",
    "start_build_from_model_number": -1,
    "test_exc_filenames": [
      "x_test.csv"
    ],
    "test_output_filenames": [
      "y_test.csv"
    ],
    "write_model_output_to_directory": "./EnergyConsumption/BuildingResults",
    "write_model_to": "./EnergyConsumption/EnergyConsumption.ednn"
  }
}

  • Place this configuration file in the same directory as the data of the test case (x_train.csv, x_test.csv, y_train.csv, y_test.csv), otherwise adjust the relative paths to the data files in the configuration file.

  • To launch the build, run the following command in the terminal (opened in the data directory, otherwise adjust the relative path to the configuration file):

neurecoDNN build ./build_configuration_file.conf

  • The build starts automatically:

Log initiated with levels: info warning error

00h00m00s info >
00h00m00s info >      _   __                ______
00h00m00s info >     / | / /__  __  _______/ ____/________
00h00m00s info >    /  |/ / _ \/ / / / ___/ __/ / ___/ __ \
00h00m00s info >   / /|  /  __/ /_/ / /  / /___/ /__/ /_/ /
00h00m00s info >  /_/ |_/\___/\__,_/_/  /_____/\___/\____/
00h00m00s info >                  === A D A G O S ===
00h00m00s info >
00h00m00s info > Version: 4.01.2591.0 Compiled with MSVC v1928  Dec  5 2022 Matlab runtime:no
00h00m00s info > OpenMP: yes
00h00m00s info > MKL: yes
00h00m00s info > Reading Dataset...
00h00m00s info > Reading data files...

To evaluate a Tabular Regression model using the executable:

 {
    "NeurEcoEvaluate": {
        "exc_filenames": [
            "x_test.csv"
        ],
        "neureco_filename": "./EnergyConsumption.ednn",
        "optional_output_reference": [
            "y_test.csv"
        ],
        "write_model_output_to_directory": "./EvaluationResults"
    }
}

  • Place this configuration file in the same directory as the data of the test case (x_train.csv, x_test.csv, y_train.csv, y_test.csv), otherwise adjust the relative paths to the data files in the configuration file

  • To launch the evaluation, run the following command in the terminal (opened in the data directory, otherwise adjust the relative path to the configuration file):

neurecoDNN evaluate ./eval_configuration_file.conf

  • The model is evaluated on the testing data in “x_test.csv”, and the results are saved in a the directory created by NeurEco: “./EvaluationResults”.

To export the Tabular Regression model using the executable (see Export NeurEco Regression model with the command line interface, embed license is required):

  • To export the model to a C format (header_file), run:

neurecoDNN exportC ./EnergyConsumption.ednn ./EnergyConsumption.h double

  • To export the model to the ONNX format, run:

neurecoDNN exportONNX ./EnergyConsumption.ednn ./EnergyConsumption.onnx float16

  • To export the model to the VBA format, run:

neurecoDNN exportVBA ./EnergyConsumption.ednn ./EnergyConsumption.onnx float

  • To export the model to the FMU format, run:

neurecoDNN exportFMU ./EnergyConsumption.ednn ./EnergyConsumption.fmu