Python tensorflow.contrib.framework.python.ops.variables.create_global_step() Examples

The following are 5 code examples of tensorflow.contrib.framework.python.ops.variables.create_global_step(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module tensorflow.contrib.framework.python.ops.variables , or try the search function .
Example #1
Source File: estimator.py    From lambda-packs with MIT License 5 votes vote down vote up 
def _infer_model(self,
                   input_fn,
                   feed_fn=None,
                   outputs=None,
                   as_iterable=True,
                   iterate_batches=False):
    # Check that model has been trained.
    checkpoint_path = saver.latest_checkpoint(self._model_dir)
    if not checkpoint_path:
      raise NotFittedError("Couldn't find trained model at %s."
                           % self._model_dir)

    with ops.Graph().as_default() as g:
      random_seed.set_random_seed(self._config.tf_random_seed)
      contrib_framework.create_global_step(g)
      features = self._get_features_from_input_fn(input_fn)
      infer_ops = self._get_predict_ops(features)
      predictions = self._filter_predictions(infer_ops.predictions, outputs)
      mon_sess = monitored_session.MonitoredSession(
          session_creator=monitored_session.ChiefSessionCreator(
              checkpoint_filename_with_path=checkpoint_path,
              scaffold=infer_ops.scaffold,
              config=self._session_config))
      if not as_iterable:
        with mon_sess:
          if not mon_sess.should_stop():
            return mon_sess.run(predictions, feed_fn() if feed_fn else None)
      else:
        return self._predict_generator(mon_sess, predictions, feed_fn,
                                       iterate_batches)
Example #2
Source File: estimator.py    From auto-alt-text-lambda-api with MIT License 5 votes vote down vote up 
def _infer_model(self,
                   input_fn,
                   feed_fn=None,
                   outputs=None,
                   as_iterable=True,
                   iterate_batches=False):
    # Check that model has been trained.
    checkpoint_path = saver.latest_checkpoint(self._model_dir)
    if not checkpoint_path:
      raise NotFittedError("Couldn't find trained model at %s."
                           % self._model_dir)

    with ops.Graph().as_default() as g:
      random_seed.set_random_seed(self._config.tf_random_seed)
      contrib_framework.create_global_step(g)
      features = self._get_features_from_input_fn(input_fn)
      infer_ops = self._call_legacy_get_predict_ops(features)
      predictions = self._filter_predictions(infer_ops.predictions, outputs)
      mon_sess = monitored_session.MonitoredSession(
          session_creator=monitored_session.ChiefSessionCreator(
              checkpoint_filename_with_path=checkpoint_path))
      if not as_iterable:
        with mon_sess:
          if not mon_sess.should_stop():
            return mon_sess.run(predictions, feed_fn() if feed_fn else None)
      else:
        return self._predict_generator(mon_sess, predictions, feed_fn,
                                       iterate_batches)
Example #3
Source File: estimator.py    From keras-lambda with MIT License 5 votes vote down vote up 
def _infer_model(self,
                   input_fn,
                   feed_fn=None,
                   outputs=None,
                   as_iterable=True,
                   iterate_batches=False):
    # Check that model has been trained.
    checkpoint_path = saver.latest_checkpoint(self._model_dir)
    if not checkpoint_path:
      raise NotFittedError("Couldn't find trained model at %s."
                           % self._model_dir)

    with ops.Graph().as_default() as g:
      random_seed.set_random_seed(self._config.tf_random_seed)
      contrib_framework.create_global_step(g)
      features = self._get_features_from_input_fn(input_fn)
      infer_ops = self._call_legacy_get_predict_ops(features)
      predictions = self._filter_predictions(infer_ops.predictions, outputs)
      mon_sess = monitored_session.MonitoredSession(
          session_creator=monitored_session.ChiefSessionCreator(
              checkpoint_filename_with_path=checkpoint_path))
      if not as_iterable:
        with mon_sess:
          if not mon_sess.should_stop():
            return mon_sess.run(predictions, feed_fn() if feed_fn else None)
      else:
        return self._predict_generator(mon_sess, predictions, feed_fn,
                                       iterate_batches)
Example #4
Source File: estimator.py    From auto-alt-text-lambda-api with MIT License 4 votes vote down vote up 
def _train_model(self, input_fn, hooks):
    all_hooks = []
    self._graph = ops.Graph()
    with self._graph.as_default() as g, g.device(self._device_fn):
      random_seed.set_random_seed(self._config.tf_random_seed)
      global_step = contrib_framework.create_global_step(g)
      features, labels = input_fn()
      self._check_inputs(features, labels)
      model_fn_ops = self._call_legacy_get_train_ops(features, labels)
      ops.add_to_collection(ops.GraphKeys.LOSSES, model_fn_ops.loss)
      all_hooks.extend([
          basic_session_run_hooks.NanTensorHook(model_fn_ops.loss),
          basic_session_run_hooks.LoggingTensorHook(
              {
                  'loss': model_fn_ops.loss,
                  'step': global_step
              },
              every_n_iter=100)
      ])
      all_hooks.extend(hooks)

      scaffold = model_fn_ops.training_scaffold or monitored_session.Scaffold()
      if not (scaffold.saver or ops.get_collection(ops.GraphKeys.SAVERS)):
        ops.add_to_collection(
            ops.GraphKeys.SAVERS,
            saver.Saver(
                sharded=True,
                max_to_keep=self._config.keep_checkpoint_max,
                defer_build=True))

      chief_hooks = []
      if (self._config.save_checkpoints_secs or
          self._config.save_checkpoints_steps):
        saver_hook_exists = any([
            isinstance(h, basic_session_run_hooks.CheckpointSaverHook)
            for h in (all_hooks + model_fn_ops.training_hooks + chief_hooks +
                      model_fn_ops.training_chief_hooks)
        ])
        if not saver_hook_exists:
          chief_hooks = [
              basic_session_run_hooks.CheckpointSaverHook(
                  self._model_dir,
                  save_secs=self._config.save_checkpoints_secs,
                  save_steps=self._config.save_checkpoints_steps,
                  scaffold=scaffold)
          ]
      with monitored_session.MonitoredTrainingSession(
          master=self._config.master,
          is_chief=self._config.is_chief,
          checkpoint_dir=self._model_dir,
          scaffold=scaffold,
          hooks=all_hooks + model_fn_ops.training_hooks,
          chief_only_hooks=chief_hooks + model_fn_ops.training_chief_hooks,
          save_checkpoint_secs=0,  # Saving is handled by a hook.
          save_summaries_steps=self._config.save_summary_steps,
          config=self.config.tf_config) as mon_sess:
        loss = None
        while not mon_sess.should_stop():
          _, loss = mon_sess.run([model_fn_ops.train_op, model_fn_ops.loss])
      summary_io.SummaryWriterCache.clear()
      return loss
Example #5
Source File: estimator.py    From keras-lambda with MIT License 4 votes vote down vote up 
def _train_model(self, input_fn, hooks):
    all_hooks = []
    self._graph = ops.Graph()
    with self._graph.as_default() as g, g.device(self._device_fn):
      random_seed.set_random_seed(self._config.tf_random_seed)
      global_step = contrib_framework.create_global_step(g)
      features, labels = input_fn()
      self._check_inputs(features, labels)
      model_fn_ops = self._call_legacy_get_train_ops(features, labels)
      ops.add_to_collection(ops.GraphKeys.LOSSES, model_fn_ops.loss)
      all_hooks.extend([
          basic_session_run_hooks.NanTensorHook(model_fn_ops.loss),
          basic_session_run_hooks.LoggingTensorHook(
              {
                  'loss': model_fn_ops.loss,
                  'step': global_step
              },
              every_n_iter=100)
      ])
      all_hooks.extend(hooks)

      scaffold = model_fn_ops.training_scaffold or monitored_session.Scaffold()
      if not (scaffold.saver or ops.get_collection(ops.GraphKeys.SAVERS)):
        ops.add_to_collection(
            ops.GraphKeys.SAVERS,
            saver.Saver(
                sharded=True,
                max_to_keep=self._config.keep_checkpoint_max,
                defer_build=True))

      chief_hooks = []
      if (self._config.save_checkpoints_secs or
          self._config.save_checkpoints_steps):
        saver_hook_exists = any([
            isinstance(h, basic_session_run_hooks.CheckpointSaverHook)
            for h in (all_hooks + model_fn_ops.training_hooks + chief_hooks +
                      model_fn_ops.training_chief_hooks)
        ])
        if not saver_hook_exists:
          chief_hooks = [
              basic_session_run_hooks.CheckpointSaverHook(
                  self._model_dir,
                  save_secs=self._config.save_checkpoints_secs,
                  save_steps=self._config.save_checkpoints_steps,
                  scaffold=scaffold)
          ]
      with monitored_session.MonitoredTrainingSession(
          master=self._config.master,
          is_chief=self._config.is_chief,
          checkpoint_dir=self._model_dir,
          scaffold=scaffold,
          hooks=all_hooks + model_fn_ops.training_hooks,
          chief_only_hooks=chief_hooks + model_fn_ops.training_chief_hooks,
          save_checkpoint_secs=0,  # Saving is handled by a hook.
          save_summaries_steps=self._config.save_summary_steps,
          config=self.config.tf_config) as mon_sess:
        loss = None
        while not mon_sess.should_stop():
          _, loss = mon_sess.run([model_fn_ops.train_op, model_fn_ops.loss])
      summary_io.SummaryWriterCache.clear()
      return loss