trainer

Primitive Trainer Interface.

Classes

Trainer

Trainer provide an interface for all trainers to inherit from.

class ashpy.trainers.trainer.Trainer(epochs, example_dim, logdir=PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/ashpy/checkouts/master/docs/source/log'), log_eval_mode=<LogEvalMode.TEST: 1>, global_step=None, metrics=None, callbacks=None)

Bases: abc.ABC

Trainer provide an interface for all trainers to inherit from.

__init__(epochs, example_dim, logdir=PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/ashpy/checkouts/master/docs/source/log'), log_eval_mode=<LogEvalMode.TEST: 1>, global_step=None, metrics=None, callbacks=None)

Primitive trainer interface. Handles model saving and restore.

Parameters:

• epochs (int) – Number of training epochs.
• example_dim (Tuple[int, int]) – Dimension of an example. In the case of GANs the example has dimension (2,1) since it’s composed by a tuple in which the first element is a tuple with 2 components and the second component is a single element. In the case of classifier the example has dimension (1, 1) since it’s composed by the example and the label.
• logdir (str) – Checkpoint and log directory.
• log_eval_mode (py:class:ashpy.modes.LogEvalMode) – to use when evaluating and logging.
• global_step (Optional[py:class:ashpy.modes.LogEvalMode]) – tf.Variable that keeps track of the training steps.
• metrics (Optional[List | Tuple [ashpy.metrics.Metric]]) – list or tuple of metrics.
• callbacks (Optional[List[ashpy.callbacks.Callback]]) – list or tuple of callbacks to handle events.

Return type:

None

_build_and_restore_models(dataset)

Build and restore a Subclassed model by firstly calling it on some data.

static _check_name_collision(objects, obj_type)

Check that all objects have unique name.

_current_epoch()

Get the current epoch using the (restored) variables.

Return type:Tensor Returns:current_epoch (tf.Tensor) – the current epoch of training.

_dataset_from_example(example, dims)

Get a dataset from a given example.

Return type:DatasetV2 Returns:The dataset containing only the example.

_generate_checkpoint_map()

Generate a human readable map of the id and type mapping in the checkpoint.

_log_metrics_and_reset()

Call for each metric the log and reset_states.

_measure_performance()

Measure performance on dataset.

_measure_performance_if_needed(example, measure_performance_freq)

Measure performance if needed.

Measure performance if self._global_step % measure_performance_freq is 0.

_on_batch_end()

Handle the end of a training batch.

Return type:None

_on_batch_start()

Handle the start of a training batch.

Return type:None

_on_epoch_end()

Handle the end of the training epoch.

Return type:None

_on_epoch_start()

Handle the start of the training epoch.

Return type:None

_on_exception()

Handle the exception.

Return type:None

_on_train_end()

Handle the end of training.

Return type:None

_on_train_start()

Handle the start of training.

Return type:None

_reduce(per_replica_tensor, reduce_op)

Reduce the input tensor in a distributed fashion, using the specified op.

_restore_or_init()

Restore or initialize the persistence layer (checkpoint).

_save()

Save the current checkpointable object status.

_update_checkpoint(ckpt_dict)

Update the checkpoint with the new checkpoint dictionary.

_update_global_batch_size(dataset, executors=None)

Set the self._global_batch_size variable where needed.

Parameters:dataset (tf.data.Dataset) – a dataset from which the batch size will be extracted.

:param executors (Union[List[ashpy.losses.executor.Executor],: ashpy.losses.executor.Executor]: a list of executor

with the property “global_batch_size”.

_validate_callbacks()

Check if every callback is an ashpy.callbacks.Callback.

_validate_metrics()

Check if every metric is an ashpy.metrics.Metric.

call(*args, **kwargs)

Execute the training process.

context

Return the training context.

Return type:Context

local_example(example, dims)

Return a local example from a distributed example.

Returns:A local example from a distributed example.

measure_metrics()

Measure the metrics.

Return type:None

model_selection()

Use the metrics to perform model selection.

Return type:None