Training Tab: Run configuration

---

Configuration file

Select the JSON file with the training parameters using the Browse button, or use the ? button to automatically load a default template.

If you use the AiDataGenByLeo trainer, the JSON format must be as follows:

{
    "general": {
        "_coment" : "Path a la carpeta base de datos donde se ubican otros folders de trabajo", 
        "data_path": "C:\\Users\\USER\\AppData\\Roaming\\MetaQuotes\\Terminal\\Common\\Files",
        "coment_prev" : "Ruta relativa a data_path, donde se ubica el folder symbol",
        "path_project" : "EasySbAi\\EURUSD",
        "_coment_1" : "Sera necesario especificar que archivo features usas para cada tipo de modelo ML estos archovs son rutas relativas a: ",
        "_coment_2" : "1. Symbol folder  | 2. Timeframe folder | 3. label_labelid folder... internamente se usa un sistema de cascada y sobreescritura del archivo",
        "_coment_3" : "El archivo debe de estar en formato .csv, si tiene el archivo .fgblc use el editor (integrao en el panel) para cargarlo y luego compilarlo a csv",
        "_coment_4" : "En caso el .fgblc este embebido en el ea recomiendo sacar ese archivo en un archivo .fgblc y ubiquelo en FeatureEditor\\My\\",
        "_coment_5" : "(path relativo al folder del panel), ahora renombrelo (en base a los nombres los csv) y luego cargarlo en el editor y posteriomente compilarlo",
        "features_pred_file": "Features\\features_model.csv",
        "features_tp_file": "Features\\features_model.csv",
        "features_sl_file": "Features\\features_model.csv",
        "file_name_idx" : "idx.txt",
        "file_name_features_ptr" : "features_ptr.txt"
    },
    "clasificacion": {
        "target_col": " salida",
        "model_name": "ModelPred",
        "num_features": 25,
        "validation_split": 0.2,
        "n_trials": 75,
        "k_folds": 5,
        "random_seed": 42,
        "hilos": 2,
        "jobs_optuna": 12,
        "final_hilos": 20,
        "data_csv_file" : "data_pred.csv"
    },
    "regresion": {
        "target_col": " salida",
        "model_name_tp": "ModelTP",
        "model_name_sl": "ModelSL",
        "num_features": 25,
        "validation_split": 0.2,
        "n_trials": 75,
        "k_folds": 5,
        "random_seed": 42,
        "hilos": 2,
        "jobs_optuna": 12,
        "final_hilos": 20,
        "_coment" : "Aqui se ubican los archivos de salida para tpy sl, el nombre de estos su ubicacion es relativa a MainFolder en este caso EasySb",
        "data_csv_file_tp": "data_tp.csv",
        "data_csv_file_sl" : "data_sl.csv"
    }
}

Description of the main fields:

general

• data_path: Absolute path to the base folder (common fies or MQL5 Files).
• path_project: The path of the specific folder symbol is specified relative to "data_path".
• features_pred_file, features_tp_file, features_sl_file: Paths relative to the symbol folder for the .csv feature files of each model. If you have a .fgblc, compile it first from the Feature Editor.
• file_name_idx: Name of the file where the indices relative to the features file will be stored.
• file_name_features_ptr: Name of the features (pointer) file where paths to the actual feature files will be stored.

clasificacion

• target_col: Name of the target column in the CSV.
• model_name: Name of the classification model.
• num_features: Number of features to select.
• validation_split: Proportion of data for validation (0.2 = 20%).
• n_trials: Number of Optuna trials for hyperparameter optimisation.
• k_folds: Number of folds for cross-validation.
• random_seed: Random seed for reproducibility.
• hilos: Threads for final training.
• jobs_optuna: Parallel threads for Optuna.
• final_hilos: Threads for the final trained model.
• data_csv_file: CSV file with training data, relative to the label_labelid folder.

regresion: Same fields as clasificacion with the difference that there are two models (model_name_tp and model_name_sl) and two data files (data_csv_file_tp and data_csv_file_sl).

If you have created your own Python trainer, use the JSON format corresponding to your implementation.

---

Other fields

• Python file: Use the Browse button to select the trainer file. .exe and .py formats are allowed.

  To run a .py directly you must configure the InpTrainingPyInterpete and InpTrainingPyPath parameters in the EA inputs. See: EA-Parameters

• Log file: Optional. Specify a .log file where all Python process logs will be written. If left empty, logs will be shown directly in the process console.

• Execution time limit: Maximum time in seconds the Python process is allowed to run. Default: 1800 seconds (30 minutes).

• Run button: Starts the training process with the current configuration. The button changes to Running... while the process is in progress.