How to Use The Tool

1. Initialize a workspace. This page uses example_project as the example workspace name. By default, only the sample dataset (ds0) is included.

   Run this in Python. On macOS, start Python with python3 if python is not available.

   import pynnlf
   
   pynnlf.init("example_project")
   

2. Fill in the input values for your forecast problem and model specification in example_project/specs/experiment.yaml.

   dataset: ds0
   forecast_horizon: fh1  # fh1 = 30 minutes ahead
   model: m6              # Linear regression
   hyperparameter: hp1    # Hyperparameter ID
   

3. Run the experiment.

   import pynnlf
   
   pynnlf.run_experiment("example_project/specs/experiment.yaml")
   

4. To skip plot generation and save storage, use the per-run override.

   import pynnlf
   
   pynnlf.run_experiment(
       "example_project/specs/experiment.yaml",
       plot_enabled=False,
   )
   

5. Run a batch of experiments (optional).

   import pynnlf
   
   pynnlf.run_experiment_batch("example_project/specs/batch.yaml")
   

   To skip plot generation for every experiment in the batch:

   import pynnlf
   
   pynnlf.run_experiment_batch(
       "example_project/specs/batch.yaml",
       plot_enabled=False,
   )
   

6. Recap multiple experiments into one CSV (optional).

   import pynnlf
   
   pynnlf.recap_experiments("example_project/experiment_result")
   

   This writes example_project/experiment_result/a1_experiment_result.csv and skips missing or malformed CSVs with warnings.

7. The tool outputs evaluation results to example_project/experiment_result/. The *_cv1_plots/ folder is created only when plot generation is enabled.

For the list of available datasets and models, how to modify model hyperparameters, how to add a model, how to add a dataset, and the API reference, see the Detailed Guide page.