ad2deb9b161610b0e9afc7a1277d7d7e6b724180,returnn/tf/network.py,LossHolder,_tf_summary,#LossHolder#,2254

Before Change

          tf_compat.v1.summary.scalar("exp_loss_%s" % name, tf.exp(self._loss_value * self._norm_factor))
        if self._network.get_config().bool("debug_unnormalized_loss_summaries", False):
          tf_compat.v1.summary.scalar("unnormalized_loss_%s" % name, self._loss_value)
        if self._network.get_config().bool("debug_objective_loss_summaries", False):
          tf_compat.v1.summary.scalar("objective_loss_%s" % name, self._loss_value_for_objective)
    if self._error_value is not None:
      if self._error_value.get_shape().ndims == 0:

After Change

          tf_compat.v1.summary.scalar("exp_loss_%s" % name, tf.exp(self._loss_value * self._norm_factor))
        if self._network.get_config().bool("debug_unnormalized_loss_summaries", False):
          tf_compat.v1.summary.scalar("unnormalized_loss_%s" % name, self._loss_value)
        if self._network.get_config().bool("debug_objective_loss_summaries", False) and self._loss_value_for_objective:
          tf_compat.v1.summary.scalar("objective_loss_%s" % name, self._loss_value_for_objective)
    if self._error_value is not None:
      if self._error_value.get_shape().ndims == 0:

Instances