Skip to main content
App version: 3.4.14

Log parameters and model configuration

To store parameters, hyperparameters, configs, or other single values, use the log_configs() function.

To organize and label the metadata in the run structure, define namespaces.

In the below example, the parameters are stored in a namespace called parameters. Inside that namespace, an attribute is created for each parameter.

from neptune_scale import Run


if __name__ == "__main__":
    run = Run(experiment_name=...)
    run.log_configs(
        {
            "parameters/learning_rate": 0.001,
            "parameters/batch_size": 64,
        },
    )

Log nested dictionaries​

To log nested data, use the flatten parameter to handle the structure:

nested_data = {
    "parameters": {
        "learning_rate": 0.001,
        "batch_size": 64,
    },
    "metrics": {
        "token_count": 76420,
        "agg": {
            "loss": 0.13,
            "acc": 0.97,
        }
    },
}

if __name__ == "__main__":
    run = Run(experiment_name=...)

    run.log_configs(data=nested_data, flatten=True)

You'll get a similar nested structure in the resulting Neptune run:

run/
|-- parameters/
    |-- learning_rate: 0.001
    |-- batch_size: 64
|-- metrics/
    |-- token_count: 76420
    |-- agg/
        |-- loss: 0.13
        |-- acc: 0.97

Cast unsupported values​

If your dictionary contains data types that Neptune doesn't support, you can enable type casting with the cast_unsupported parameter:

data = {
    "metrics": {
        "token_count": 76420,
        "agg": {
            "loss": None,
            "acc": None,
        }
    },
    "some_list": [1, "test", None],
}

if __name__ == "__main__":
    run = Run(experiment_name=...)

    run.log_configs(
        data=data,
        flatten=True,
        cast_unsupported=True,
    )

Unsupported values are cast to strings in the Neptune run structure:

run/
|-- metrics/
    |-- token_count: 76420
    |-- agg/
        |-- loss: ""
        |-- acc: ""
|-- some_list: '[1, "test", None]'

View configs in the app​

You can find your logged configs in the Attributes section of the run.

In dashboards and reports, you can use the single value widget to display a value from one selected run.

To compare configs or other single values for multiple runs at once, use the following: