.. _guide:

Configuration File
===================

The config file drives the evaluation workflow. There is a default version 
provided inisde the main WeatherGenerator repository `here`_.

.. _here: https://github.com/ecmwf/WeatherGenerator/blob/main/config/evaluate/eval_config.yml

The following is not an exhaustive list (there are more features being added, for a preview check out the ``develop`` branch).

.. code-block::

    #optional: if commented out all is taken care of by the default settings
    # NB. global options apply to all run_ids 
    #global_plotting_options:
    #  image_format : "png" #options: "png", "pdf", "svg", "eps", "jpg" ..
    #  dpi_val : 300
    #  fps: 2
    #  ERA5:
    #    marker_size: 2
    #    scale_marker_size: 1
    #    marker: "o"
    #    # alpha: 0.5
    #    2t: 
    #      vmin: 250
    #      vmax: 300
    #    10u:
    #      vmin: -40
    #      vmax: 40

    evaluation:
        metrics  : ["rmse", "mae"]
        regions: ["global", "nhem"]
        summary_plots : true
        summary_dir: "./plots/"
        plot_ensemble: "members" #supported: false, "std", "minmax", "members"
        plot_score_maps: false #plot scores on a 2D maps. it slows down score computation 
        print_summary: false #print out score values on screen. it can be verbose
        log_scale: false
        add_grid: false
        score_cards: false
        bar_plots: false
        num_processes: 0  #options: int, "auto", 0 means no parallelism (default)
        # baseline: "ar40mckx"

    run_ids :
        ar40mckx:
            label: "pretrained model ar40mckx"
            results_base_dir : "./results/"
            mini_epoch: 0
            rank: 0
            streams:
                ERA5:
                    channels: ["2t", "10u"] #, "10v", "z_500", "t_850", "u_850", "v_850", "q_850", ]
                    evaluation:
                        forecast_step: "all"
                        sample: "all"
                        ensemble: "all" #supported: "all", "mean", [0,1,2]
                    plotting:
                        sample: [1, 3]
                        forecast_step: [1,3, 2]
                        ensemble: [0,2,5] #supported: "all", "mean", [0,1,2]
                        plot_maps: true
                        plot_target: false
                        plot_histograms: true
                        plot_animations: true
                CERRA:
                    channels: ["z_500", "t_850", "u_850"] #, "blah"]
                    evaluation:
                        forecast_step: "all"
                        sample: "all"
                    plotting:
                        sample: [2, 3, 0]
                        forecast_step: [1,3, 4, 5]
                        plot_maps: true
                        plot_target: false
                        plot_histograms: true
                        plot_`animations: true
                
        c8g5katp:
            label: "2 steps window"
            results_base_dir : "./results/"
            mini_epoch: 0
            rank: 0
            streams:
                ERA5:
                    channels: ["2t", "10u", "10v"] #, "10v", "z_500", "t_850", "u_850", "v_850", "q_850", ]
                    evaluation:
                        forecast_step: "all"
                        sample: "all"
                        ensemble: "mean" 
                    plotting:
                        sample: [1, 3]
                        forecast_step: [1,3, 2]
                        ensemble: "mean" 
                        plot_maps: true
                        plot_target: false
                        plot_histograms: true
                        plot_animations: true
    
Further details on config options can be found:

.. toctree::
    :maxdepth: 1

    global
    regions
    metrics
    csv
    plotting