Source code for ashpy.contexts.context

# Copyright 2019 Zuru Tech HK Limited. All Rights Reserved.
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.

Primitive Context Interface.

``Contexts`` are checkpointable (subclassed from :py:class:`tf.train.Checkpoint`)
collections of variable encapsulated in a Python Class as a way to seamlessly
handle information transfer.

from typing import Optional, Tuple

import tensorflow as tf
from ashpy.metrics import Metric
from ashpy.modes import LogEvalMode

[docs]class Context: """:py:class:`ashpy.contexts.Context` provide an interface for all contexts."""
[docs] def __init__( self, metrics: Tuple[Metric] = None, dataset: = None, log_eval_mode: LogEvalMode = LogEvalMode.TEST, global_step=tf.Variable(0, name="global_step", trainable=False, dtype=tf.int64), checkpoint: tf.train.Checkpoint = None, ) -> None: """ Initialize the Context. Args: metrics (:obj:`tuple` of (:py:class:`ashpy.metrics.metric.Metric`)): List of :py:class:`ashpy.metrics.metric.Metric` objects. dataset (:py:class:``): The dataset to use, that contains everything needed to use the model in this context. log_eval_mode (:py:class:`ashpy.modes.LogEvalMode`): Models' mode to use when evaluating and logging. global_step (:py:class:`tf.Variable`): Keeps track of the training steps. checkpoint (:py:class:`tf.train.Checkpoint`): Checkpoint to use to keep track of models status. """ self._distribute_strategy = tf.distribute.get_strategy() self._metrics = metrics if metrics else () self._dataset = dataset self._log_eval_mode = log_eval_mode self._global_step = global_step self._checkpoint = checkpoint self._exception: Optional[Exception] = None self._current_batch: Optional[tf.Tensor] = None
@property def log_eval_mode(self) -> LogEvalMode: """ Retrieve model(s) mode. Returns: :py:class:`ashpy.modes.LogEvalMode`. """ return self._log_eval_mode @property def dataset(self) -> """ Retrieve the dataset. Returns: :py:class:`` the current dataset """ return self._dataset @dataset.setter def dataset(self, _dataset: """ Set the dataset. Args: _dataset (:py:class:``): dataset to set """ self._dataset = _dataset @property def metrics(self) -> Tuple[Metric]: """ Retrieve the metrics. Returns: :obj:`tuple` of (:py:class:`ashpy.metrics.metric.Metric`). """ return self._metrics @property def global_step(self) -> tf.Variable: """ Retrieve the global_step. Returns: :py:class:`tf.Variable`. """ return self._global_step @property def exception(self) -> Optional[Exception]: """Return the exception.""" return self._exception @exception.setter def exception(self, exception: Optional[Exception]) -> None: """Set the exception.""" self._exception = exception @property def current_batch(self) -> Optional[tf.Tensor]: """Return the current batch.""" return self._current_batch @current_batch.setter def current_batch(self, _current_batch: Optional[tf.Tensor]) -> None: """Set the current_batch.""" self._current_batch = _current_batch