Output files¶
Monte Carlo class will generate at least four different type of files on the path ./logs/montecarlo/<year>/<month>/<day>/run_<time>/
run_data.csv
: Raw results with theloss
,epoch
,model
, etc.<model_name>_statistical_result.csv
: One file is generated for each model in the monte carlo simulation. It contains the statistical results per epoch of the model.models_details.json
: A full detailed description of each model to be trained.run/iteration<iteration>_model<model index and name>/<model_name>_results_fit.csv
: Inside therun
folder there is information of the result for each model at each iteration.
There are many other optional files that can be controled using the output_config dictionary variable of booleans:
output_config = {
'plot_all': False,
'confusion_matrix': False,
'excel_summary': True,
'summary_of_run': True,
'tensorboard': False,
'save_weights': False,
'safety_checkpoints': False
}
A complementary file test_results.csv
can be generated if test_data
is passed run()
(None
by default).
Usage example:
monte_carlo = MonteCarlo()
monte_carlo.output_config['plot_all'] = True # Tell monte carlo to save the plots
Variable | Generated File(s) | Description |
---|---|---|
‘excel_summary’ | .logs/*.xlsx | Adds a row to an excel file with information of the monte carlo run |
./logs/montecarlo/<year>/<month>/<day>/run_<time>/run_data.csv | Always generated. All monte carlo raw results. | |
./logs/montecarlo/<year>/<month>/<day>/run_<time>/<model_name>_statistical_result.csv | Always generated. Statistical results per model. | |
./logs/montecarlo/<year>/<month>/<day>/run_<time>/models_details.json | Always generated. A full detailed description of each model to be trained. | |
‘plot_all’ | ./logs/montecarlo/<year>/<month>/<day>/run_<time>/plots/ | Plots generated using MonteCarloAnalyzer.do_all() |
‘confusion_matrix’ | ./logs/montecarlo/<year>/<month>/<day>/run_<time>/<model_name>_confusion_matrix.csv | Generates a confusion matrix file per model |
‘summary_of_run’ | ./logs/montecarlo/<year>/<month>/<day>/run_<time>/run_summary.txt | Generates a user friendly monte carlo text summary |
‘safety_checkpoints’ | ./logs/montecarlo/<year>/<month>/<day>/run_<time>/run_data.csv | Creates the run_data.csv as data is obtained not to lose information if an unexpected exit happens. |
./logs/montecarlo/<year>/<month>/<day>/run_<time>/test_results.csv | This is generated if a test_data is passed to the run() method of MonteCarlo. |
|
‘tensorboard’ | ./logs/montecarlo/<year>/<month>/<day>/run_<time>/tensorboard/ | Saves tensorboard results like graph, loss and accuracy. |