Quickstart: Tabular Compression with the command line interface#

NeurEco executable for Tabular models (Regression, Compression, Classification) is called NeurEcoDNN. It can be called directly from a terminal / powershell after a full installation of NeurEco.

To build a NeurEco Regression model, run the following command in the terminal:

neurecoDNN build path/to/build/configuration/file/build.conf

The skeleton of a configuration file required to build NeurEco Regression model, here build.conf, looks as follows (for the test case Heaviside). Its fields should be filled according to the problem at hand.

    {
"neurecoDNN_build": {
    "DevSettings": {
        "compressor_decompressor_ratio": 1,
        "final_learning": true,
        "initial_beta_reg": 0.1,
        "valid_percentage": 33.33
    },
    "UserSettings": {
        "gpu_id": 0,
        "use_gpu": false
    },
    "build_compress": true,
    "checkpoint_address": "./Heaviside/Heaviside.checkpoint",
    "classification": false,
    "compress_tolerance": 0.05,
    "exc_filenames": [
        "x_train.csv"
    ],
    "freeze_structure": false,
    "input_normalization": {
        "normalize_per_feature": false,
        "scale_type": "max",
        "shift_type": "none"
    },
    "minimum_compression_coefficients": 1,
    "model_output_format": "csv",
    "output_normalization": {
        "normalize_per_feature": false,
        "scale_type": "max",
        "shift_type": "none"
    },
    "resume": false,
    "starting_from_checkpoint_address": "",
    "starting_from_model_id": -1,
    "test_exc_filenames": [
        "x_test.csv"
    ],
    "write_compression_model_to": "./Heaviside/compressionHeaviside.ednn",
    "write_decompression_model_to": "./Heaviside/decompressionHeaviside.ednn",
    "write_model_to": "./Heaviside/Heaviside.ednn"
    }
    }

To perform an evaluation, run the following command in the terminal:

neurecoDNN evaluate path/to/evaluation/configuration/file/eval.conf

The skeleton of an evaluation configuration file, here eval.conf, looks as follows (for the test case Heaviside). Its fields should be filled according to the problem at hand.

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

Note

For detailed documentation on evaluate, see Evaluate NeurEco Compression model in the command line interface

To export the model to the chosen format, run one of the following commands:

neurecoDNN exportC ./Heaviside/Heaviside.ednn ./Heaviside.h double
neurecoDNN exportONNX ./Heaviside/Heaviside.ednn ./Heaviside.onnx float16
neurecoDNN exportVBA ./Heaviside/Heaviside.ednn ./Heaviside.onnx float
neurecoDNN exportFMU ./Heaviside/Heaviside.ednn ./Heaviside.fmu

Export to these formats requires embed license.

Note

For detailed documentation on Tabular Compression with the command line interface, see Tabular Compression with the command line interface.