The main use cases are documentet as the Machine Learning Pipeline. This page documents a few additional features.
Training of individual models¶
Make a config TOML file, specifying the model and the hyperparamers. An example is provided as Test/experiment01.toml
python -m droulette.model --config Test/experiment01.tomlData is generated in the directory specified in the TOML file.
In particular, the trained model is stored (model.pth) to be able to make additional analyses
Plot training development¶
To plot the development of the loss function over epochs, the following
module can be used. In batch, it creates a plot file training.svg
from the training_log.csv files in each directory.
python -m droulette.diagnostic --alldirs _test/experiment???It can also be run on individual log files, without the --alldirs option.