src.utils.logging package

Submodules

src.utils.logging.base_logger module

This file implements the base logger which is an interface for loggers.

class src.utils.logging.base_logger.BaseLogger

Bases: ABC

This class is the interface for all loggers.

abstract log_end(data: Dict[str, Any]) → None

Logging function that is called after training for logging training statistics. For example, it can be used to parse tensorflows history object after it is returned after training.

Parameters:

data – The dictionary containing training data and metrics.

abstract log_epoch(data: Dict[str, Any]) → None

Logging function that is called every epoch for logging epoch statistics. For example, it can be used to save the epoch’s loss and metrics to the overall logging data.

Parameters:

data – The dictionary containing epoch data and metrics.

abstract log_start(data: Dict[str, Any]) → None

Logging function that is called before training for logging training parameters and metrics. For example, it can be used to save initialization parameters or configuration details.

Parameters:

data – The dictionary containing data and config.

save_logs(folder: str) → None

This function saves the gathered data in a json file for review and plotting the statistics later.

Parameters:

folder – The folder to store the statistics in.

src.utils.logging.pytorch_logger module

Implement a logger for tensorflow keras models.

class src.utils.logging.pytorch_logger.TorchLogger

Bases: BaseLogger

Implements a logger for pytorch models that receives relevant values every epoch and creates the logs from them.

log_end(data: Dict[str, Any]) → None

Log additional data at the end of the training if necessary.

Parameters:

data – Data dictionary containing important logs.

log_epoch(data: Dict[str, Any]) → None

Epoch logging method. Pass a loss, val_loss, acc and val_acc every epoch.

Parameters:

data – Dictionary with data to log.

log_start(data: Dict[str, Any]) → None

Log arbitrary data like training parameters etc.

Parameters:

data – Data dictionary containing important data to log.

src.utils.logging.standard_logger module

This file implements the base logger which is an interface for loggers.

class src.utils.logging.standard_logger.StandardLogger

Bases: BaseLogger

This class is a standard logger that logs random dictionaries.

log_end(data: Dict[str, Any]) → None

Logging function that is called after training for logging training statistics.

Parameters:

data – The dictionary containing training data and metrics.

log_epoch(data: Dict[str, Any]) → None

Logging function that is called every epoch for logging epoch statistics. For example, it can be used to save the epoch’s loss and metrics to the overall logging data.

Parameters:

data – The dictionary containing epoch data and metrics.

log_start(data: Dict[str, Any]) → None

Logging function that is called before training for logging training parameters and metrics. For example, it can be used to save initialization parameters or configuration details.

Parameters:

data – The dictionary containing data and config.

src.utils.logging.tensorflow_logger module

Implement a logger for tensorflow keras models.

class src.utils.logging.tensorflow_logger.KerasLogger

Bases: BaseLogger

Implements a logger for keras models that uses tensorflow’s history object and extracts the logs from it.

log_end(data: Dict[str, Any]) → None

Log all the gathered training logs from tensorflow and keras.

Parameters:

data – Data dictionary containing important logs. history: We expect a history object to be present here.

log_epoch(data: Dict[str, Any]) → None

Empty epoch logging method. Tensorflow logs all epoch logs itself and returns it in a history object. Therefore, no logging per epoch is required.

Parameters:

data – Dictionary with data to log.

log_start(data: Dict[str, Any]) → None

Log arbitrary data like training parameters etc.

Parameters:

data – Data dictionary containing important data to log.

Module contents

Logging functionality is implemented in this package.