Exceptions¶
- exception substrafl.exceptions.CriterionReductionError¶
Bases:
Exception
The criterion reduction must be set to ‘mean’ to use the Newton-Raphson strategy.
- exception substrafl.exceptions.DampingFactorValueError¶
Bases:
Exception
The damping factor must be greater than 0 and less than or equal to 1
- exception substrafl.exceptions.DatasetSignatureError¶
Bases:
Exception
The __init__() function on the given torch Dataset must have x, y and is_inference as parameters.
- exception substrafl.exceptions.DatasetTypeError¶
Bases:
Exception
The given torch Dataset should be a torch.utils.data.Dataset object and not an instance of it.
Bases:
Exception
The shared_states is empty. Ensure that the train method of the algorithm returns a StrategySharedState object.
- exception substrafl.exceptions.IncompatibleAlgoStrategyError¶
Bases:
Exception
This algo is not compatible with this strategy.
- exception substrafl.exceptions.IndexGeneratorSampleNoneError¶
Bases:
Exception
Try to use the index generator without setting the number of samples.
- exception substrafl.exceptions.IndexGeneratorUpdateError¶
Bases:
Exception
The index generator has not been updated properly.
- exception substrafl.exceptions.KeyMetadataError¶
Bases:
Exception
substrafl_version
,substra_version
andsubstratools_version
keys can’t be added to the experiment metadata.
- exception substrafl.exceptions.LenMetadataError¶
Bases:
Exception
Too long additional metadata passed to the execute_experiment function to be shown on the Substra WebApp.
- exception substrafl.exceptions.LoadAlgoFileNotFoundError¶
Bases:
Exception
When using the
load_algo()
, the given folder must contains the following files: algo.tar.gz, metadata.json, the file entered in thelocal_state_file
key of the dictionary.
- exception substrafl.exceptions.LoadAlgoLocalDependencyError¶
Bases:
Exception
When using the
load_algo()
, all dependencies from the local input folder should be install by the user.
- exception substrafl.exceptions.LoadAlgoMetadataError¶
Bases:
Exception
When using the
load_algo()
, the metadata.json file within the folder given as input must contain alocal_state_file
- exception substrafl.exceptions.MultipleTrainTaskError¶
Bases:
Exception
When using the
download_algo_files()
function in remote mode. The experiment to get the algo files from can’t have multiple TrainDataNodes hosted on the same organization.
- exception substrafl.exceptions.NegativeHessianMatrixError¶
Bases:
Exception
Hessian matrix is not positive semi-definite, the problem is not convex
- exception substrafl.exceptions.NumUpdatesValueError¶
Bases:
Exception
The num_update has been set to an non-authorize value.
- exception substrafl.exceptions.OptimizerValueError¶
Bases:
Exception
The optimizer value is incompatible with the _local_train function.
- exception substrafl.exceptions.ScaffoldLearningRateError¶
Bases:
Exception
When using the
TorchScaffoldAlgo
, a learning rate must be passed to the optimizer.
Bases:
Exception
Shared states error
- exception substrafl.exceptions.SubstraToolsDeprecationWarning¶
Bases:
DeprecationWarning
The substratools version used is deprecated.
- exception substrafl.exceptions.TorchScaffoldAlgoParametersUpdateError¶
Bases:
Exception
When using
TorchScaffoldAlgo
,_scaffold_parameters_update()
method must be called once for each update within the_local_train()
method.
- exception substrafl.exceptions.TrainTaskNotFoundError¶
Bases:
Exception
When using the
download_algo_files()
function. The provided compute plan must contain a composite train tuple:hosted by the worker associated to the given client
tagged with the given round_idx
- exception substrafl.exceptions.UnfinishedTrainTaskError¶
Bases:
Exception
When using the
download_algo_files()
function. The task to get the algo files from shall be in statusSTATUS_DONE
.