Question

在培训和评估代码部分。（{Complete code参见此处）：

<dependencies>
    <!-- https://mvnrepository.com/artifact/org.json/json -->
    <dependency>
        <groupId>org.json</groupId>
        <artifactId>json</artifactId>
        <version>20180130</version>
    </dependency>
    </dependencies>

JSONObject cannot be resolved to a type
The import org.json cannot be resolved

我有2个问题：

  train_hooks = hooks_helper.get_train_hooks(
      flags_obj.hooks, model_dir=flags_obj.model_dir,
      batch_size=flags_obj.batch_size, tensors_to_log=tensors_to_log)

  # Train and evaluate the model every `flags.epochs_between_evals` epochs.
  for n in range(flags_obj.train_epochs // flags_obj.epochs_between_evals):
    model.train(input_fn=train_input_fn, hooks=train_hooks)

    results = model.evaluate(input_fn=eval_input_fn)

    # Display evaluation metrics
    tf.logging.info('Results at epoch %d / %d',
                    (n + 1) * flags_obj.epochs_between_evals,
                    flags_obj.train_epochs)
    tf.logging.info('-' * 60)


    for key in sorted(results):
      tf.logging.info('%s: %s' % (key, results[key]))


    benchmark_logger.log_evaluation_result(results)


    if early_stop and model_helpers.past_stop_threshold(
        flags_obj.stop_threshold, results['accuracy']):
      break

默认设置为2，但是如何在训练后每2个时期评估1个时代？显然每个epochs_between_evals之后有一个model.evaluate()，不是吗？
也许model.train()控制着train_hooks中的内容每隔两个纪元中断一次，但是model.train()的参数不包含train_hooks或num_epochs。我有点困惑。
epochs_between_evals的实际作用是什么？当我检查源代码时，似乎只是获得评估输出的转发途径。还有其他工作吗？

张量流的广泛和深入模型中的评估代码理解

0 个答案: