Package jidenn
This package contains the code for the JIDENN project. It contains all the necessary code to train and evaluate the models used for jet identification.
The process starts with the jidenn.data
package, which contains the code to load the data,
preprocess it, and create the datasets used for training and evaluation. It is recommended to
convert the ROOT file to saved tf.data.Dataset
objects, as this is much faster than loading directly.
On top of that it provides an easy way to create own training inputs, with the
jidenn.data.TrainInput
module.
The models are defined in the jidenn.models
package. There are both, models that take the
jet constituents as input, and models that take the high-level jet variables already present in the
ROOT file as input.
The model creation is done with the jidenn.model_builders
module. This module contains
multiple facades for the model creation, separating individual steps to make it easier to
extend the code at any point.
All configuration is done with the hydra
package. The configuration files are located in the
jidenn/config
folder. The jidenn.config
module contains the dataclasses that connect the
*.yaml
files to the code.
Evaluation is done with the jidenn.evaluation
module. This module contains the code to evaluate
custom metrics, and to plot the results.
Expand source code
"""
This package contains the code for the JIDENN project.
It contains all the necessary code to train and evaluate the models used for jet identification.
The process starts with the `jidenn.data` package, which contains the code to load the data,
preprocess it, and create the datasets used for training and evaluation. It is recommended to
convert the ROOT file to saved `tf.data.Dataset` objects, as this is much faster than loading directly.
On top of that it provides an easy way to create own training inputs, with the
`jidenn.data.TrainInput` module.
The models are defined in the `jidenn.models` package. There are both, models that take the
jet constituents as input, and models that take the high-level jet variables already present in the
ROOT file as input.
The model creation is done with the `jidenn.model_builders` module. This module contains
multiple facades for the model creation, separating individual steps to make it easier to
extend the code at any point.
All configuration is done with the `hydra` package. The configuration files are located in the
`jidenn/config` folder. The `jidenn.config` module contains the dataclasses that connect the
`*.yaml` files to the code.
Evaluation is done with the `jidenn.evaluation` module. This module contains the code to evaluate
custom metrics, and to plot the results.
"""
Sub-modules
jidenn.config
-
This module to fully configure the data preparation, training and evaulation. The configuration is done using the hydra package …
jidenn.const
jidenn.data
-
This module contains all the necessary toolkit to: …
jidenn.evaluation
-
Module for evaluating models. The evalutaion consists of: calculating metrics and plotting the results. Custom metrics are defined in …
jidenn.model_builders
-
Module for building models. It separates the model building process into two parts: model initialization and model building with output layer, …
jidenn.models
-
This module contains all the models used for the experiments on jet identification. They can be basicaly clustered into three categories: …
jidenn.preprocess