optuna_transformers
See also
Full documentation with examples can be found here: documentation page
Integration of Optuna and Transformers.
- class hpoflow.optuna_transformers.OptunaMLflowCallback(trial, log_training_args=True, log_model_config=True)[source]
Bases:
TrainerCallbackIntegration of Optuna and Transformers.
Class based on
transformers.TrainerCallback; integrates with OptunaMLflow to send the logs toMLflowandOptunaduring model training.Constructor.
- Parameters:
trial (OptunaMLflow) – The OptunaMLflow object.
log_training_args (bool) – Whether to log all Transformers TrainingArguments as MLflow params.
log_model_config (bool) – Whether to log the Transformers model config as MLflow params.
- on_log(args, state, control, logs, model=None, **kwargs)[source]
Event called after logging the last logs.
Log all metrics from Transformers logs as MLflow metrics at the appropriate step.
- on_train_begin(args, state, control, model=None, **kwargs)[source]
Event called at the beginning of training.
Call setup if not yet initialized.
- on_train_end(args, state, control, **kwargs)[source]
Event called at the end of training.
Log the training output as MLflow artifacts if logging artifacts is enabled.
- Parameters:
args (TrainingArguments) –
state (TrainerState) –
control (TrainerControl) –
- setup(args, state, model)[source]
Setup the optional MLflow integration.
You can set the environment variable
HF_MLFLOW_LOG_ARTIFACTS. It is to usemlflow.log_artifacts()to log artifacts. This only makes sense if logging to a remote server, e.g. s3 or GCS. If set toTrueor1, will copy whatever is in TrainerArgument’s output_dir to the local or remote artifact storage. Using it without a remote storage will just copy the files to your artifact location.