f302eba7375d8da3477816a705c7ce628f9a55a8,dytb/trainer/Trainer.py,Trainer,train,#Trainer#,42

Before Change

            // General validation summary
            metric_value_ = tf.placeholder(tf.float32, shape=())
            metric_summary = tf.summary.scalar(
                self._model.evaluator.metric["name"], metric_value_)

            // read collection after that every op added its own
            // summaries in the train_summaries collection
            train_summaries = tf.summary.merge(

After Change

            train_metric = None
            for metric in self._model.evaluator.metrics:
                if metric["model_selection"] is not None:
                    train_metric = metric
                else:
                    tf_log(
                        tf.summary.scalar(metric["name"], metric["fn"](
                            predictions, targets)))

            if train_metric is None:
                print(
                    "Please specify a metric in the evaluator with "model_selection" not None"
                )

Instances