 在 TensorFlow.org 上查看
|
 在 Google Colab 中运行
|
 在 GitHub 上查看源码
|
 下载笔记本
|
本文档介绍了 tf.estimator——一种高级 TensorFlow API。Estimators 封装了以下操作
- 训练
- 评估 (Evaluation)
- 预测
- 导出以供服务
TensorFlow 实现了几种预制的 Estimator。自定义 Estimator 仍然受到支持，但主要是作为向后兼容的措施。新代码不应使用自定义 Estimator。所有的 Estimator（无论是预制的还是自定义的）都是基于 tf.estimator.Estimator 类的类。
如需快速示例，请尝试 Estimator 教程。有关 API 设计的概述，请查阅白皮书。
设置
pip install -U tensorflow_datasetsimport tempfile
import os
import tensorflow as tf
import tensorflow_datasets as tfds
2024-01-24 02:20:44.732508: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered 2024-01-24 02:20:44.732557: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered 2024-01-24 02:20:44.734001: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
优势
与 tf.keras.Model 类似，estimator 是一种模型层面的抽象。tf.estimator 提供了一些目前仍在为 tf.keras 开发中的功能。这些功能包括：
- 基于参数服务器的训练
- 与 TFX 的完全集成
Estimators 的功能
Estimators 提供以下益处：
- 您可以在本地主机或分布式多服务器环境上运行基于 Estimator 的模型，而无需更改模型。此外，您可以在 CPU、GPU 或 TPU 上运行基于 Estimator 的模型，而无需重新编写代码。
- Estimators 提供了一个安全的分布式训练循环，用于控制何时以及如何：
- 加载数据
- 处理异常
- 创建检查点文件并从故障中恢复
- 保存用于 TensorBoard 的摘要 (summaries)
在使用 Estimators 编写应用程序时，必须将数据输入流水线与模型分离开来。这种分离简化了使用不同数据集进行实验的过程。
使用预制 Estimator
预制 Estimator 使您能够在比基础 TensorFlow API 更高的概念水平上工作。您无需再担心创建计算图或会话，因为 Estimator 会为您处理所有“管道”工作。此外，预制 Estimator 让您只需极少的代码更改即可试验不同的模型架构。例如，tf.estimator.DNNClassifier 就是一个预制 Estimator 类，用于训练基于密集前馈神经网络的分类模型。
依赖于预制 Estimator 的 TensorFlow 程序通常由以下四个步骤组成：
1. 编写输入函数 (input functions)
例如，您可以创建一个函数来导入训练集，另一个函数来导入测试集。Estimator 要求其输入格式化为一对对象：
- 一个字典，其中键是特征名称，值是包含相应特征数据的张量（或稀疏张量）
- 一个包含一个或多个标签的张量
input_fn 应返回一个能产生上述格式对的 tf.data.Dataset。
例如，以下代码从泰坦尼克号数据集的 train.csv 文件构建了一个 tf.data.Dataset：
def train_input_fn():
titanic_file = tf.keras.utils.get_file("train.csv", "https://storage.googleapis.com/tf-datasets/titanic/train.csv")
titanic = tf.data.experimental.make_csv_dataset(
titanic_file, batch_size=32,
label_name="survived")
titanic_batches = (
titanic.cache().repeat().shuffle(500)
.prefetch(tf.data.AUTOTUNE))
return titanic_batches
input_fn 在 tf.Graph 中执行，也可以直接返回包含图张量的 (features_dics, labels) 对，但在简单的常量返回场景之外，这种做法容易出错。
2. 定义特征列 (feature columns)。
每个 tf.feature_column 标识一个特征名称、其类型以及任何输入预处理。
例如，以下代码片段创建了三个特征列。
- 第一个直接将
age特征用作浮点输入。 - 第二个将
class特征用作分类输入。 - 第三个将
embark_town用作分类输入，但使用“哈希技巧”(hashing trick) 以避免枚举所有选项，并设置选项数量。
有关更多信息，请查看特征列教程。
age = tf.feature_column.numeric_column('age')
cls = tf.feature_column.categorical_column_with_vocabulary_list('class', ['First', 'Second', 'Third'])
embark = tf.feature_column.categorical_column_with_hash_bucket('embark_town', 32)
WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/4100412726.py:1: numeric_column (from tensorflow.python.feature_column.feature_column_v2) is deprecated and will be removed in a future version. Instructions for updating: Use Keras preprocessing layers instead, either directly or via the `tf.keras.utils.FeatureSpace` utility. Each of `tf.feature_column.*` has a functional equivalent in `tf.keras.layers` for feature preprocessing when training a Keras model. WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/4100412726.py:2: categorical_column_with_vocabulary_list (from tensorflow.python.feature_column.feature_column_v2) is deprecated and will be removed in a future version. Instructions for updating: Use Keras preprocessing layers instead, either directly or via the `tf.keras.utils.FeatureSpace` utility. Each of `tf.feature_column.*` has a functional equivalent in `tf.keras.layers` for feature preprocessing when training a Keras model. WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/4100412726.py:3: categorical_column_with_hash_bucket (from tensorflow.python.feature_column.feature_column_v2) is deprecated and will be removed in a future version. Instructions for updating: Use Keras preprocessing layers instead, either directly or via the `tf.keras.utils.FeatureSpace` utility. Each of `tf.feature_column.*` has a functional equivalent in `tf.keras.layers` for feature preprocessing when training a Keras model.
3. 实例化相关的预制 Estimator。
例如，这是名为 LinearClassifier 的预制 Estimator 的实例化示例：
model_dir = tempfile.mkdtemp()
model = tf.estimator.LinearClassifier(
model_dir=model_dir,
feature_columns=[embark, cls, age],
n_classes=2
)
WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/1904628810.py:2: LinearClassifierV2.__init__ (from tensorflow_estimator.python.estimator.canned.linear) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/head/head_utils.py:54: BinaryClassHead.__init__ (from tensorflow_estimator.python.estimator.head.binary_class_head) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/canned/linear.py:944: Estimator.__init__ (from tensorflow_estimator.python.estimator.estimator) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1844: RunConfig.__init__ (from tensorflow_estimator.python.estimator.run_config) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
INFO:tensorflow:Using default config.
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpbt9n791j', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
欲了解更多信息，请参阅线性分类器教程。
4. 调用训练、评估或推理方法。
所有 Estimator 都提供 train、evaluate 和 predict 方法。
model = model.train(input_fn=train_input_fn, steps=100)
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:385: StopAtStepHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. Downloading data from https://storage.googleapis.com/tf-datasets/titanic/train.csv 30874/30874 [==============================] - 0s 0us/step INFO:tensorflow:Calling model_fn. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/optimizers/legacy/ftrl.py:173: calling Constant.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version. Instructions for updating: Call initializer instance with the dtype argument instead of passing it to the constructor WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/model_fn.py:250: EstimatorSpec.__new__ (from tensorflow_estimator.python.estimator.model_fn) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Done calling model_fn. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1416: NanTensorHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1419: LoggingTensorHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/basic_session_run_hooks.py:232: SecondOrStepTimer.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1456: CheckpointSaverHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Create CheckpointSaverHook. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/monitored_session.py:579: StepCounterHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/monitored_session.py:586: SummarySaverHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Graph was finalized. INFO:tensorflow:Running local_init_op. INFO:tensorflow:Done running local_init_op. INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0... INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpbt9n791j/model.ckpt. INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0... WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/monitored_session.py:1455: SessionRunArgs.__new__ (from tensorflow.python.training.session_run_hook) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/monitored_session.py:1454: SessionRunContext.__init__ (from tensorflow.python.training.session_run_hook) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/monitored_session.py:1474: SessionRunValues.__new__ (from tensorflow.python.training.session_run_hook) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:loss = 0.6931472, step = 0 INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 100... INFO:tensorflow:Saving checkpoints for 100 into /tmpfs/tmp/tmpbt9n791j/model.ckpt. INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100... INFO:tensorflow:Loss for final step: 0.59910536. 2024-01-24 02:20:52.016569: W tensorflow/core/kernels/data/cache_dataset_ops.cc:858] The calling iterator did not fully read the dataset being cached. In order to avoid unexpected truncation of the dataset, the partially cached contents of the dataset will be discarded. This can happen if you have an input pipeline similar to `dataset.cache().take(k).repeat()`. You should use `dataset.take(k).cache().repeat()` instead.
result = model.evaluate(train_input_fn, steps=10)
for key, value in result.items():
print(key, ":", value)
INFO:tensorflow:Calling model_fn. INFO:tensorflow:Done calling model_fn. INFO:tensorflow:Starting evaluation at 2024-01-24T02:20:52 WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/evaluation.py:260: FinalOpsHook.__init__ (from tensorflow.python.training.basic_session_run_hooks) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Graph was finalized. INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpbt9n791j/model.ckpt-100 INFO:tensorflow:Running local_init_op. INFO:tensorflow:Done running local_init_op. INFO:tensorflow:Evaluation [1/10] INFO:tensorflow:Evaluation [2/10] INFO:tensorflow:Evaluation [3/10] INFO:tensorflow:Evaluation [4/10] INFO:tensorflow:Evaluation [5/10] INFO:tensorflow:Evaluation [6/10] INFO:tensorflow:Evaluation [7/10] INFO:tensorflow:Evaluation [8/10] INFO:tensorflow:Evaluation [9/10] INFO:tensorflow:Evaluation [10/10] INFO:tensorflow:Inference Time : 3.02373s INFO:tensorflow:Finished evaluation at 2024-01-24-02:20:55 INFO:tensorflow:Saving dict for global step 100: accuracy = 0.715625, accuracy_baseline = 0.615625, auc = 0.7548389, auc_precision_recall = 0.66771877, average_loss = 0.56671846, global_step = 100, label/mean = 0.384375, loss = 0.56671846, precision = 0.7, prediction/mean = 0.37144834, recall = 0.45528457 INFO:tensorflow:Saving 'checkpoint_path' summary for global step 100: /tmpfs/tmp/tmpbt9n791j/model.ckpt-100 accuracy : 0.715625 accuracy_baseline : 0.615625 auc : 0.7548389 auc_precision_recall : 0.66771877 average_loss : 0.56671846 label/mean : 0.384375 loss : 0.56671846 precision : 0.7 prediction/mean : 0.37144834 recall : 0.45528457 global_step : 100 2024-01-24 02:20:55.868009: W tensorflow/core/kernels/data/cache_dataset_ops.cc:858] The calling iterator did not fully read the dataset being cached. In order to avoid unexpected truncation of the dataset, the partially cached contents of the dataset will be discarded. This can happen if you have an input pipeline similar to `dataset.cache().take(k).repeat()`. You should use `dataset.take(k).cache().repeat()` instead.
for pred in model.predict(train_input_fn):
for key, value in pred.items():
print(key, ":", value)
break
INFO:tensorflow:Calling model_fn. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/head/base_head.py:786: ClassificationOutput.__init__ (from tensorflow.python.saved_model.model_utils.export_output) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/head/binary_class_head.py:561: RegressionOutput.__init__ (from tensorflow.python.saved_model.model_utils.export_output) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/head/binary_class_head.py:563: PredictOutput.__init__ (from tensorflow.python.saved_model.model_utils.export_output) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Done calling model_fn. INFO:tensorflow:Graph was finalized. INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpbt9n791j/model.ckpt-100 INFO:tensorflow:Running local_init_op. INFO:tensorflow:Done running local_init_op. logits : [-0.6829183] logistic : [0.33561027] probabilities : [0.6643897 0.33561027] class_ids : [0] classes : [b'0'] all_class_ids : [0 1] all_classes : [b'0' b'1'] 2024-01-24 02:20:56.898963: W tensorflow/core/kernels/data/cache_dataset_ops.cc:858] The calling iterator did not fully read the dataset being cached. In order to avoid unexpected truncation of the dataset, the partially cached contents of the dataset will be discarded. This can happen if you have an input pipeline similar to `dataset.cache().take(k).repeat()`. You should use `dataset.take(k).cache().repeat()` instead.
预制 Estimator 的优势
预制 Estimator 编码了最佳实践，提供了以下益处：
- 确定计算图的不同部分应在何处运行的最佳实践，以及在单机或集群上实现策略的最佳实践。
- 事件（摘要）编写和通用有用摘要的最佳实践。
如果不使用预制 Estimator，您必须自己实现上述功能。
自定义 Estimator
每个 Estimator（无论是预制的还是自定义的）的核心都是其模型函数，即 model_fn，这是一个构建训练、评估和预测计算图的方法。当使用预制 Estimator 时，他人已为您实现了模型函数。当依赖自定义 Estimator 时，必须自己编写模型函数。
从 Keras 模型创建 Estimator
您可以使用 tf.keras.estimator.model_to_estimator 将现有的 Keras 模型转换为 Estimator。如果您想现代化模型代码，但训练流水线仍需要 Estimator，这会很有帮助。
实例化一个 Keras MobileNet V2 模型，并使用优化器、损失函数和指标编译该模型以进行训练。
keras_mobilenet_v2 = tf.keras.applications.MobileNetV2(
input_shape=(160, 160, 3), include_top=False)
keras_mobilenet_v2.trainable = False
estimator_model = tf.keras.Sequential([
keras_mobilenet_v2,
tf.keras.layers.GlobalAveragePooling2D(),
tf.keras.layers.Dense(1)
])
# Compile the model
estimator_model.compile(
optimizer='adam',
loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
metrics=['accuracy'])
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_160_no_top.h5 9406464/9406464 [==============================] - 0s 0us/step
从编译后的 Keras 模型创建 Estimator。Keras 模型的初始模型状态将保留在创建的 Estimator 中。
est_mobilenet_v2 = tf.keras.estimator.model_to_estimator(keras_model=estimator_model)
INFO:tensorflow:Using default config.
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmpjcucleiz
INFO:tensorflow:Using the Keras model provided.
WARNING:absl:You are using `tf.keras.optimizers.experimental.Optimizer` in TF estimator, which only supports `tf.keras.optimizers.legacy.Optimizer`. Automatically converting your optimizer to `tf.keras.optimizers.legacy.Optimizer`.
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/backend.py:452: UserWarning: `tf.keras.backend.set_learning_phase` is deprecated and will be removed after 2020-10-11. To update it, simply pass a True/False value to the `training` argument of the `__call__` method of your layer or model.
warnings.warn(
2024-01-24 02:21:01.520653: W tensorflow/c/c_api.cc:305] Operation '{name:'block_16_project_BN/gamma/Assign' id:2164 op device:{requested: '', assigned: ''} def:{ { {node block_16_project_BN/gamma/Assign} } = AssignVariableOp[_has_manual_control_dependencies=true, dtype=DT_FLOAT, validate_shape=false](block_16_project_BN/gamma, block_16_project_BN/gamma/Initializer/ones)} }' was changed by setting attribute after it was run by a session. This mutation will have no effect, and will trigger an error in the future. Either don't modify nodes after running them or create a new session.
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpjcucleiz', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpjcucleiz', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:2404: WarmStartSettings.__new__ (from tensorflow_estimator.python.estimator.estimator) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:2404: WarmStartSettings.__new__ (from tensorflow_estimator.python.estimator.estimator) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
像对待任何其他 Estimator 一样对待派生的 Estimator。
IMG_SIZE = 160 # All images will be resized to 160x160
def preprocess(image, label):
image = tf.cast(image, tf.float32)
image = (image/127.5) - 1
image = tf.image.resize(image, (IMG_SIZE, IMG_SIZE))
return image, label
def train_input_fn(batch_size):
data = tfds.load('cats_vs_dogs', as_supervised=True)
train_data = data['train']
train_data = train_data.map(preprocess).shuffle(500).batch(batch_size)
return train_data
要进行训练，调用 Estimator 的 train 函数。
est_mobilenet_v2.train(input_fn=lambda: train_input_fn(32), steps=50)
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Warm-starting with WarmStartSettings: WarmStartSettings(ckpt_to_initialize_from='/tmpfs/tmp/tmpjcucleiz/keras/keras_model.ckpt', vars_to_warm_start='.*', var_name_to_vocab_info={}, var_name_to_prev_var_name={})
INFO:tensorflow:Warm-starting with WarmStartSettings: WarmStartSettings(ckpt_to_initialize_from='/tmpfs/tmp/tmpjcucleiz/keras/keras_model.ckpt', vars_to_warm_start='.*', var_name_to_vocab_info={}, var_name_to_prev_var_name={})
INFO:tensorflow:Warm-starting from: /tmpfs/tmp/tmpjcucleiz/keras/keras_model.ckpt
INFO:tensorflow:Warm-starting from: /tmpfs/tmp/tmpjcucleiz/keras/keras_model.ckpt
INFO:tensorflow:Warm-starting variables only in TRAINABLE_VARIABLES.
INFO:tensorflow:Warm-starting variables only in TRAINABLE_VARIABLES.
INFO:tensorflow:Warm-started 158 variables.
INFO:tensorflow:Warm-started 158 variables.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpjcucleiz/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpjcucleiz/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:loss = 0.6796611, step = 0
INFO:tensorflow:loss = 0.6796611, step = 0
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 50...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 50...
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmpjcucleiz/model.ckpt.
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmpjcucleiz/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 50...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 50...
INFO:tensorflow:Loss for final step: 0.69532585.
INFO:tensorflow:Loss for final step: 0.69532585.
<tensorflow_estimator.python.estimator.estimator.EstimatorV2 at 0x7f29400d3670>
同样，要进行评估，调用 Estimator 的 evaluate 函数。
est_mobilenet_v2.evaluate(input_fn=lambda: train_input_fn(32), steps=10)
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/engine/training_v1.py:2335: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
updates = self.state_updates
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Starting evaluation at 2024-01-24T02:21:19
INFO:tensorflow:Starting evaluation at 2024-01-24T02:21:19
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpjcucleiz/model.ckpt-50
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpjcucleiz/model.ckpt-50
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Evaluation [1/10]
INFO:tensorflow:Evaluation [1/10]
INFO:tensorflow:Evaluation [2/10]
INFO:tensorflow:Evaluation [2/10]
INFO:tensorflow:Evaluation [3/10]
INFO:tensorflow:Evaluation [3/10]
INFO:tensorflow:Evaluation [4/10]
INFO:tensorflow:Evaluation [4/10]
INFO:tensorflow:Evaluation [5/10]
INFO:tensorflow:Evaluation [5/10]
INFO:tensorflow:Evaluation [6/10]
INFO:tensorflow:Evaluation [6/10]
INFO:tensorflow:Evaluation [7/10]
INFO:tensorflow:Evaluation [7/10]
INFO:tensorflow:Evaluation [8/10]
INFO:tensorflow:Evaluation [8/10]
INFO:tensorflow:Evaluation [9/10]
INFO:tensorflow:Evaluation [9/10]
INFO:tensorflow:Evaluation [10/10]
INFO:tensorflow:Evaluation [10/10]
INFO:tensorflow:Inference Time : 2.81526s
INFO:tensorflow:Inference Time : 2.81526s
INFO:tensorflow:Finished evaluation at 2024-01-24-02:21:22
INFO:tensorflow:Finished evaluation at 2024-01-24-02:21:22
INFO:tensorflow:Saving dict for global step 50: accuracy = 0.54375, global_step = 50, loss = 0.6558426
INFO:tensorflow:Saving dict for global step 50: accuracy = 0.54375, global_step = 50, loss = 0.6558426
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50: /tmpfs/tmp/tmpjcucleiz/model.ckpt-50
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50: /tmpfs/tmp/tmpjcucleiz/model.ckpt-50
{'accuracy': 0.54375, 'loss': 0.6558426, 'global_step': 50}
更多详情，请参阅 tf.keras.estimator.model_to_estimator 的文档。
使用 Estimator 保存基于对象的检查点
Estimator 默认使用变量名保存检查点，而不是 Checkpoint 指南中描述的对象图。tf.train.Checkpoint 可以读取基于名称的检查点，但当将模型的部分移出 Estimator 的 model_fn 时，变量名可能会发生变化。为了保持前向兼容性，保存基于对象的检查点可以更容易地在 Estimator 内部训练模型，然后在外部使用它。
import tensorflow.compat.v1 as tf_compat
def toy_dataset():
inputs = tf.range(10.)[:, None]
labels = inputs * 5. + tf.range(5.)[None, :]
return tf.data.Dataset.from_tensor_slices(
dict(x=inputs, y=labels)).repeat().batch(2)
class Net(tf.keras.Model):
"""A simple linear model."""
def __init__(self):
super(Net, self).__init__()
self.l1 = tf.keras.layers.Dense(5)
def call(self, x):
return self.l1(x)
def model_fn(features, labels, mode):
net = Net()
opt = tf.keras.optimizers.Adam(0.1)
ckpt = tf.train.Checkpoint(step=tf_compat.train.get_global_step(),
optimizer=opt, net=net)
with tf.GradientTape() as tape:
output = net(features['x'])
loss = tf.reduce_mean(tf.abs(output - features['y']))
variables = net.trainable_variables
gradients = tape.gradient(loss, variables)
return tf.estimator.EstimatorSpec(
mode,
loss=loss,
train_op=tf.group(opt.apply_gradients(zip(gradients, variables)),
ckpt.step.assign_add(1)),
# Tell the Estimator to save "ckpt" in an object-based format.
scaffold=tf_compat.train.Scaffold(saver=ckpt))
tf.keras.backend.clear_session()
est = tf.estimator.Estimator(model_fn, './tf_estimator_example/')
est.train(toy_dataset, steps=10)
INFO:tensorflow:Using default config.
INFO:tensorflow:Using default config.
INFO:tensorflow:Using config: {'_model_dir': './tf_estimator_example/', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Using config: {'_model_dir': './tf_estimator_example/', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Saving checkpoints for 0 into ./tf_estimator_example/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into ./tf_estimator_example/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:loss = 4.4765882, step = 1
INFO:tensorflow:loss = 4.4765882, step = 1
WARNING:tensorflow:It seems that global step (tf.train.get_global_step) has not been increased. Current value (could be stable): 4 vs previous value: 4. You could increase the global step by passing tf.train.get_global_step() to Optimizer.apply_gradients or Optimizer.minimize.
WARNING:tensorflow:It seems that global step (tf.train.get_global_step) has not been increased. Current value (could be stable): 4 vs previous value: 4. You could increase the global step by passing tf.train.get_global_step() to Optimizer.apply_gradients or Optimizer.minimize.
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10...
INFO:tensorflow:Saving checkpoints for 10 into ./tf_estimator_example/model.ckpt.
INFO:tensorflow:Saving checkpoints for 10 into ./tf_estimator_example/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10...
INFO:tensorflow:Loss for final step: 44.10201.
INFO:tensorflow:Loss for final step: 44.10201.
<tensorflow_estimator.python.estimator.estimator.EstimatorV2 at 0x7f29407836a0>
tf.train.Checkpoint 随后可以从 model_dir 加载 Estimator 的检查点。
opt = tf.keras.optimizers.Adam(0.1)
net = Net()
ckpt = tf.train.Checkpoint(
step=tf.Variable(1, dtype=tf.int64), optimizer=opt, net=net)
ckpt.restore(tf.train.latest_checkpoint('./tf_estimator_example/'))
ckpt.step.numpy() # From est.train(..., steps=10)
10
来自 Estimator 的 SavedModel
Estimator 通过 tf.Estimator.export_saved_model 导出 SavedModel。
input_column = tf.feature_column.numeric_column("x")
estimator = tf.estimator.LinearClassifier(feature_columns=[input_column])
def input_fn():
return tf.data.Dataset.from_tensor_slices(
({"x": [1., 2., 3., 4.]}, [1, 1, 0, 0])).repeat(200).shuffle(64).batch(16)
estimator.train(input_fn)
INFO:tensorflow:Using default config.
INFO:tensorflow:Using default config.
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmps_vspwiv
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmps_vspwiv
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmps_vspwiv', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmps_vspwiv', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmps_vspwiv/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmps_vspwiv/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:loss = 0.6931472, step = 0
INFO:tensorflow:loss = 0.6931472, step = 0
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 50...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 50...
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmps_vspwiv/model.ckpt.
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmps_vspwiv/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 50...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 50...
INFO:tensorflow:Loss for final step: 0.42848104.
INFO:tensorflow:Loss for final step: 0.42848104.
<tensorflow_estimator.python.estimator.canned.linear.LinearClassifierV2 at 0x7f2940787640>
要保存 Estimator，需要创建一个 serving_input_receiver。此函数构建 tf.Graph 的一部分，用于解析 SavedModel 接收到的原始数据。
tf.estimator.export 模块包含有助于构建这些 receiver 的函数。
以下代码基于 feature_columns 构建了一个接收器，它接收序列化的 tf.Example 协议缓冲区，这在 tf-serving 中经常使用。
tmpdir = tempfile.mkdtemp()
serving_input_fn = tf.estimator.export.build_parsing_serving_input_receiver_fn(
tf.feature_column.make_parse_example_spec([input_column]))
estimator_base_path = os.path.join(tmpdir, 'from_estimator')
estimator_path = estimator.export_saved_model(estimator_base_path, serving_input_fn)
WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/121230551.py:4: make_parse_example_spec_v2 (from tensorflow.python.feature_column.feature_column_v2) is deprecated and will be removed in a future version. Instructions for updating: Use Keras preprocessing layers instead, either directly or via the `tf.keras.utils.FeatureSpace` utility. Each of `tf.feature_column.*` has a functional equivalent in `tf.keras.layers` for feature preprocessing when training a Keras model. WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/121230551.py:4: make_parse_example_spec_v2 (from tensorflow.python.feature_column.feature_column_v2) is deprecated and will be removed in a future version. Instructions for updating: Use Keras preprocessing layers instead, either directly or via the `tf.keras.utils.FeatureSpace` utility. Each of `tf.feature_column.*` has a functional equivalent in `tf.keras.layers` for feature preprocessing when training a Keras model. WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/121230551.py:3: build_parsing_serving_input_receiver_fn (from tensorflow_estimator.python.estimator.export.export) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/121230551.py:3: build_parsing_serving_input_receiver_fn (from tensorflow_estimator.python.estimator.export.export) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/export/export.py:312: ServingInputReceiver.__new__ (from tensorflow_estimator.python.estimator.export.export) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/export/export.py:312: ServingInputReceiver.__new__ (from tensorflow_estimator.python.estimator.export.export) is deprecated and will be removed in a future version. Instructions for updating: Use tf.keras instead. INFO:tensorflow:Calling model_fn. INFO:tensorflow:Calling model_fn. INFO:tensorflow:Done calling model_fn. INFO:tensorflow:Done calling model_fn. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/saved_model/signature_def_utils_impl.py:168: build_tensor_info (from tensorflow.python.saved_model.utils_impl) is deprecated and will be removed in a future version. Instructions for updating: This API was designed for TensorFlow v1. See https://tensorflowcn.cn/guide/migrate for instructions on how to migrate your code to TensorFlow v2. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/saved_model/signature_def_utils_impl.py:168: build_tensor_info (from tensorflow.python.saved_model.utils_impl) is deprecated and will be removed in a future version. Instructions for updating: This API was designed for TensorFlow v1. See https://tensorflowcn.cn/guide/migrate for instructions on how to migrate your code to TensorFlow v2. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/saved_model/model_utils/export_utils.py:83: get_tensor_from_tensor_info (from tensorflow.python.saved_model.utils_impl) is deprecated and will be removed in a future version. Instructions for updating: This API was designed for TensorFlow v1. See https://tensorflowcn.cn/guide/migrate for instructions on how to migrate your code to TensorFlow v2. WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/saved_model/model_utils/export_utils.py:83: get_tensor_from_tensor_info (from tensorflow.python.saved_model.utils_impl) is deprecated and will be removed in a future version. Instructions for updating: This API was designed for TensorFlow v1. See https://tensorflowcn.cn/guide/migrate for instructions on how to migrate your code to TensorFlow v2. INFO:tensorflow:Signatures INCLUDED in export for Classify: ['serving_default', 'classification'] INFO:tensorflow:Signatures INCLUDED in export for Classify: ['serving_default', 'classification'] INFO:tensorflow:Signatures INCLUDED in export for Regress: ['regression'] INFO:tensorflow:Signatures INCLUDED in export for Regress: ['regression'] INFO:tensorflow:Signatures INCLUDED in export for Predict: ['predict'] INFO:tensorflow:Signatures INCLUDED in export for Predict: ['predict'] INFO:tensorflow:Signatures INCLUDED in export for Train: None INFO:tensorflow:Signatures INCLUDED in export for Train: None INFO:tensorflow:Signatures INCLUDED in export for Eval: None INFO:tensorflow:Signatures INCLUDED in export for Eval: None INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmps_vspwiv/model.ckpt-50 INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmps_vspwiv/model.ckpt-50 INFO:tensorflow:Assets added to graph. INFO:tensorflow:Assets added to graph. INFO:tensorflow:No assets to write. INFO:tensorflow:No assets to write. INFO:tensorflow:SavedModel written to: /tmpfs/tmp/tmpvsrxtilj/from_estimator/temp-1706062885/saved_model.pb INFO:tensorflow:SavedModel written to: /tmpfs/tmp/tmpvsrxtilj/from_estimator/temp-1706062885/saved_model.pb
您也可以从 Python 加载并运行该模型：
imported = tf.saved_model.load(estimator_path)
def predict(x):
example = tf.train.Example()
example.features.feature["x"].float_list.value.extend([x])
return imported.signatures["predict"](
examples=tf.constant([example.SerializeToString()]))
print(predict(1.5))
print(predict(3.5))
{'logistic': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.576523]], dtype=float32)>, 'probabilities': <tf.Tensor: shape=(1, 2), dtype=float32, numpy=array([[0.42347696, 0.576523 ]], dtype=float32)>, 'class_ids': <tf.Tensor: shape=(1, 1), dtype=int64, numpy=array([[1]])>, 'logits': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.30851614]], dtype=float32)>, 'classes': <tf.Tensor: shape=(1, 1), dtype=string, numpy=array([[b'1']], dtype=object)>, 'all_class_ids': <tf.Tensor: shape=(1, 2), dtype=int32, numpy=array([[0, 1]], dtype=int32)>, 'all_classes': <tf.Tensor: shape=(1, 2), dtype=string, numpy=array([[b'0', b'1']], dtype=object)>}
{'logistic': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.2338471]], dtype=float32)>, 'probabilities': <tf.Tensor: shape=(1, 2), dtype=float32, numpy=array([[0.7661529 , 0.23384713]], dtype=float32)>, 'class_ids': <tf.Tensor: shape=(1, 1), dtype=int64, numpy=array([[0]])>, 'logits': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[-1.1867142]], dtype=float32)>, 'classes': <tf.Tensor: shape=(1, 1), dtype=string, numpy=array([[b'0']], dtype=object)>, 'all_class_ids': <tf.Tensor: shape=(1, 2), dtype=int32, numpy=array([[0, 1]], dtype=int32)>, 'all_classes': <tf.Tensor: shape=(1, 2), dtype=string, numpy=array([[b'0', b'1']], dtype=object)>}
tf.estimator.export.build_raw_serving_input_receiver_fn 允许您创建接收原始张量而不是 tf.train.Example 的输入函数。
在 Estimator 中使用 tf.distribute.Strategy（支持有限）
tf.estimator 是一种分布式训练 TensorFlow API，最初支持异步参数服务器方法。tf.estimator 现在支持 tf.distribute.Strategy。如果您正在使用 tf.estimator，只需极少代码修改即可切换到分布式训练。通过这种方式，Estimator 用户现在可以在多个 GPU 和多个 Worker 上进行同步分布式训练，并使用 TPU。但是，Estimator 中的这种支持是有限的。有关更多详细信息，请参阅下方的当前支持的内容部分。
在 Estimator 中使用 tf.distribute.Strategy 与 Keras 情况略有不同。您无需使用 strategy.scope，而是将策略对象传入 Estimator 的 RunConfig 中。
您可以参考分布式训练指南以获取更多信息。
以下代码片段展示了如何结合使用预制 Estimator LinearRegressor 和 MirroredStrategy：
mirrored_strategy = tf.distribute.MirroredStrategy()
config = tf.estimator.RunConfig(
train_distribute=mirrored_strategy, eval_distribute=mirrored_strategy)
regressor = tf.estimator.LinearRegressor(
feature_columns=[tf.feature_column.numeric_column('feats')],
optimizer='SGD',
config=config)
INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:GPU:0', '/job:localhost/replica:0/task:0/device:GPU:1', '/job:localhost/replica:0/task:0/device:GPU:2', '/job:localhost/replica:0/task:0/device:GPU:3')
INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:GPU:0', '/job:localhost/replica:0/task:0/device:GPU:1', '/job:localhost/replica:0/task:0/device:GPU:2', '/job:localhost/replica:0/task:0/device:GPU:3')
INFO:tensorflow:Initializing RunConfig with distribution strategies.
INFO:tensorflow:Initializing RunConfig with distribution strategies.
INFO:tensorflow:Not using Distribute Coordinator.
INFO:tensorflow:Not using Distribute Coordinator.
WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/4277059788.py:4: LinearRegressorV2.__init__ (from tensorflow_estimator.python.estimator.canned.linear) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/tmp/ipykernel_9073/4277059788.py:4: LinearRegressorV2.__init__ (from tensorflow_estimator.python.estimator.canned.linear) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/canned/linear.py:1344: RegressionHead.__init__ (from tensorflow_estimator.python.estimator.head.regression_head) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/canned/linear.py:1344: RegressionHead.__init__ (from tensorflow_estimator.python.estimator.head.regression_head) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.keras instead.
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmpjagdxixr
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmpjagdxixr
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpjagdxixr', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f28cc70eaf0>, '_device_fn': None, '_protocol': None, '_eval_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f28cc70eaf0>, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1, '_distribute_coordinator_mode': None}
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpjagdxixr', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
rewrite_options {
meta_optimizer_iterations: ONE
}
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f28cc70eaf0>, '_device_fn': None, '_protocol': None, '_eval_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f28cc70eaf0>, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_checkpoint_save_graph_def': True, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1, '_distribute_coordinator_mode': None}
在这里，您使用的是预制 Estimator，但相同的代码也适用于自定义 Estimator。train_distribute 决定如何分配训练，而 eval_distribute 决定如何分配评估。这与 Keras 的另一处不同点在于，Keras 中训练和评估使用相同的策略。
现在，您可以使用输入函数来训练和评估此 Estimator：
def input_fn():
dataset = tf.data.Dataset.from_tensors(({"feats":[1.]}, [1.]))
return dataset.repeat(1000).batch(10)
regressor.train(input_fn=input_fn, steps=10)
regressor.evaluate(input_fn=input_fn, steps=10)
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1246: StrategyBase.configure (from tensorflow.python.distribute.distribute_lib) is deprecated and will be removed in a future version.
Instructions for updating:
use `update_config_proto` instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/estimator.py:1246: StrategyBase.configure (from tensorflow.python.distribute.distribute_lib) is deprecated and will be removed in a future version.
Instructions for updating:
use `update_config_proto` instead.
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/data/ops/dataset_ops.py:462: UserWarning: To make it possible to preserve tf.data options across serialization boundaries, their implementation has moved to be part of the TensorFlow graph. As a consequence, the options value is in general no longer known at graph construction time. Invoking this method in graph mode retains the legacy behavior of the original implementation, but note that the returned value might not reflect the actual value of the options.
warnings.warn("To make it possible to preserve tf.data options across "
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Collective all_reduce tensors: 2 all_reduces, num_devices = 4, group_size = 4, implementation = CommunicationImplementation.NCCL, num_packs = 1
INFO:tensorflow:Collective all_reduce tensors: 2 all_reduces, num_devices = 4, group_size = 4, implementation = CommunicationImplementation.NCCL, num_packs = 1
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Create CheckpointSaverHook.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/util.py:95: DistributedIteratorV1.initialize (from tensorflow.python.distribute.v1.input_lib) is deprecated and will be removed in a future version.
Instructions for updating:
Use the iterator's `initializer` property instead.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow_estimator/python/estimator/util.py:95: DistributedIteratorV1.initialize (from tensorflow.python.distribute.v1.input_lib) is deprecated and will be removed in a future version.
Instructions for updating:
Use the iterator's `initializer` property instead.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpjagdxixr/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpjagdxixr/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
2024-01-24 02:21:31.119820: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:31.121097: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:31.125859: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:31.126357: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:31.137091: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:31.137578: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:31.143860: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:31.144344: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
INFO:tensorflow:loss = 4.0, step = 0
INFO:tensorflow:loss = 4.0, step = 0
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10...
INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10...
INFO:tensorflow:Saving checkpoints for 10 into /tmpfs/tmp/tmpjagdxixr/model.ckpt.
INFO:tensorflow:Saving checkpoints for 10 into /tmpfs/tmp/tmpjagdxixr/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10...
INFO:tensorflow:Loss for final step: 1.1510792e-12.
INFO:tensorflow:Loss for final step: 1.1510792e-12.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Reduce to /replica:0/task:0/device:CPU:0 then broadcast to ('/replica:0/task:0/device:CPU:0',).
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Starting evaluation at 2024-01-24T02:21:35
INFO:tensorflow:Starting evaluation at 2024-01-24T02:21:35
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpjagdxixr/model.ckpt-10
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpjagdxixr/model.ckpt-10
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
2024-01-24 02:21:36.170934: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:36.172154: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:36.176484: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:36.176957: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:36.182511: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:36.182980: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2024-01-24 02:21:36.196737: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2024-01-24 02:21:36.197227: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
INFO:tensorflow:Evaluation [1/10]
INFO:tensorflow:Evaluation [1/10]
INFO:tensorflow:Evaluation [2/10]
INFO:tensorflow:Evaluation [2/10]
INFO:tensorflow:Evaluation [3/10]
INFO:tensorflow:Evaluation [3/10]
INFO:tensorflow:Evaluation [4/10]
INFO:tensorflow:Evaluation [4/10]
INFO:tensorflow:Evaluation [5/10]
INFO:tensorflow:Evaluation [5/10]
INFO:tensorflow:Evaluation [6/10]
INFO:tensorflow:Evaluation [6/10]
INFO:tensorflow:Evaluation [7/10]
INFO:tensorflow:Evaluation [7/10]
INFO:tensorflow:Evaluation [8/10]
INFO:tensorflow:Evaluation [8/10]
INFO:tensorflow:Evaluation [9/10]
INFO:tensorflow:Evaluation [9/10]
INFO:tensorflow:Evaluation [10/10]
INFO:tensorflow:Evaluation [10/10]
INFO:tensorflow:Inference Time : 1.83647s
INFO:tensorflow:Inference Time : 1.83647s
INFO:tensorflow:Finished evaluation at 2024-01-24-02:21:37
INFO:tensorflow:Finished evaluation at 2024-01-24-02:21:37
INFO:tensorflow:Saving dict for global step 10: average_loss = 1.4210855e-14, global_step = 10, label/mean = 1.0, loss = 1.4210855e-14, prediction/mean = 0.99999994
INFO:tensorflow:Saving dict for global step 10: average_loss = 1.4210855e-14, global_step = 10, label/mean = 1.0, loss = 1.4210855e-14, prediction/mean = 0.99999994
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 10: /tmpfs/tmp/tmpjagdxixr/model.ckpt-10
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 10: /tmpfs/tmp/tmpjagdxixr/model.ckpt-10
{'average_loss': 1.4210855e-14,
'label/mean': 1.0,
'loss': 1.4210855e-14,
'prediction/mean': 0.99999994,
'global_step': 10}
此处需要强调 Estimator 与 Keras 之间的另一个区别是输入处理。在 Keras 中，数据集的每个批次都会在多个副本之间自动拆分。然而，在 Estimator 中，您不会进行自动批次拆分，也不会在不同的 Worker 之间自动切分数据。您可以完全控制数据在 Worker 和设备之间的分发方式，并且必须提供 input_fn 来指定如何分发数据。
您的 input_fn 在每个 Worker 上调用一次，从而为每个 Worker 提供一个数据集。然后，从该数据集中取出一个批次喂给该 Worker 上的一个副本，从而为 1 个 Worker 上的 N 个副本消耗 N 个批次。换句话说，input_fn 返回的数据集应提供大小为 PER_REPLICA_BATCH_SIZE 的批次。步骤的全局批次大小可以计算为 PER_REPLICA_BATCH_SIZE * strategy.num_replicas_in_sync。
在执行多 Worker 训练时，您应该将数据拆分到各个 Worker 上，或者在每个 Worker 上使用随机种子进行洗牌。您可以在使用 Estimator 进行多 Worker 训练教程中查看如何执行此操作的示例。
类似地，您也可以使用多 Worker 和参数服务器策略。代码保持不变，但您需要使用 tf.estimator.train_and_evaluate，并为集群中运行的每个二进制文件设置 TF_CONFIG 环境变量。
当前支持的内容
使用除 TPUStrategy 之外的所有策略进行 Estimator 训练的支持有限。基本训练和评估应该可行，但许多高级功能（例如 v1.train.Scaffold）则不行。这种集成中可能还存在许多错误，并且没有积极改进此支持的计划（重点在于 Keras 和自定义训练循环支持）。如果可能，您应该优先使用 tf.distribute 配合这些 API。
| 训练 API | MirroredStrategy | TPUStrategy | MultiWorkerMirroredStrategy | CentralStorageStrategy | ParameterServerStrategy |
|---|---|---|---|---|---|
| Estimator API | 支持有限 | 不支持 | 支持有限 | 支持有限 | 支持有限 |
示例和教程
以下是一些展示如何将各种策略与 Estimator 配合使用的端到端示例：
- 使用 Estimator 进行多 Worker 训练教程展示了如何使用 MNIST 数据集上的
MultiWorkerMirroredStrategy进行多 Worker 训练。 - 一个在
tensorflow/ecosystem中使用 Kubernetes 模板运行带有分发策略的多 Worker 训练的端到端示例。它从一个 Keras 模型开始，并使用tf.keras.estimator.model_to_estimatorAPI 将其转换为 Estimator。 - 官方 ResNet50 模型，可以使用
MirroredStrategy或MultiWorkerMirroredStrategy进行训练。