FEDOT Command Line Interface (CLI)

FEDOT API can be called via console without writing python code. All API parameters can be marked as flags described in application Help. Prediction saves as a CSV file for future use.

For correct console application run, there should be Python environment with installed FEDOT package with all dependencies.

Pay attention that your FEDOT working project can differ from the package version installed in the environment! For setting master version as a package in environment download it through pip from GitHub with command:

pip install git+https://github.com/aimclub/FEDOT.git

Start using

The main executing script running the application is fedot_cli.py so first, there is a need to navigate to them:

cd {path_to_fedot}/fedot/api

To get a list of possible flags and their descriptions, a help call is provided:

python fedot_cli.py --help

The result of execution is presented below:

--problem PROBLEM     The name of modelling problem to solve:
                      classification;
                      regression;
                      ts_forecasting;
                      clustering
--train TRAIN         Path to train data file
--test TEST           Path to test data file
--preset PRESET       Name of preset for model building:
                      light;
                      light_steady_state;
                      ultra_light;
                      ultra_steady_state;
                      ts;
                      gpu
--timeout TIMEOUT     Time for model design (in minutes)
--seed SEED           Value for fixed random seed
--target TARGET       Name of target variable in data
--depth DEPTH         Composer parameter: max depth of the pipeline
--arity ARITY         Composer parameter: max arity of the pipeline nodes
--popsize POPSIZE     Composer parameter: population size
--gen_num GEN_NUM     Composer parameter: number of generations
--opers [OPERS [OPERS ...]]
                      Composer parameter: model names to use
--tuning TUNING       Composer parameter: 1 - with tuning, 0 - without tuning
--cv_folds CV_FOLDS   Composer parameter: Number of folds for cross-validation
--hist_path HIST_PATH
                      Composer parameter: Name of the folder for composing history
--for_len FOR_LEN     Time Series Forecasting parameter: forecast length

Examples of using (.bat files)

Examples of usage can be presented as .bat files for console execution. These files are located at /examples/cli_application folder. There the templates of parameters for different problems decision are presented.

The string below helps to run classification problem decision from the console:

python --problem classification --train ../../test/data/simple_classification.csv --test ../../test/data/simple_classification.csv  --target Y --timeout 0.1