 TensorFlow.org에서 보기 |
 Google Colab에서 실행 |
 GitHub에서 소스 보기 |
 노트북 다운로드 |
경고: Estimator는 새 코드에 권장되지 않습니다. Estimator는
v1.Session스타일 코드를 실행하며, 이 코드는 올바르게 작성하기가 좀 더 어렵고 특히 TF 2 코드와 결합할 경우 예기치 않게 작동할 수 있습니다. Estimator는 호환성 보장이 적용되지만 보안 취약점 외에는 수정 사항이 제공되지 않습니다. 자세한 내용은 마이그레이션 가이드를 참조하세요.
이 문서는 높은 수준의 TensorFlow API인 tf.estimator를 소개합니다. Estimator는 다음 동작을 캡슐화합니다.
- 훈련
- 평가
- 예측
- 처리를 위한 내보내기
TensorFlow는 사전 작성된 여러 estimator를 구현합니다. 사용자 정의 estimator도 여전히 지원되지만, 주로 하위 호환성을 위해 사용됩니다. 새 코드에는 사용자 정의 estimator를 사용하지 않아야 합니다. 사전 작성되거나 사용자가 만든 모든 estimator는 tf.estimator.Estimator 클래스를 기반으로 하는 클래스입니다.
간단한 예를 보려면 Estimator 튜토리얼을 참조하세요. API 설계에 대한 개요는 백서를 참조하세요.
설정
pip install -U tensorflow_datasetsimport tempfile
import os
import tensorflow as tf
import tensorflow_datasets as tfds
2022-12-14 21:57:20.391732: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory 2022-12-14 21:57:20.391822: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory 2022-12-14 21:57:20.391832: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
이점
tf.keras.Model과 유사하게, estimator는 모델 레벨 추상화입니다. tf.estimator는 tf.keras용으로 현재 개발 중인 일부 기능을 제공합니다. 다음과 같습니다.
- 매개변수 서버 기반 훈련
- 완전한 TFX 통합
Estimator 기능
Estimator는 다음과 같은 이점을 제공합니다.
- 모델을 변경하지 않고 로컬 호스트 또는 분산 다중 서버 환경에서 Estimator 기반 모델을 실행할 수 있습니다. 또한, 모델을 다시 코딩하지 않고 CPU, GPU 또는 TPU에서 Estimator 기반 모델을 실행할 수 있습니다.
- Estimator는 다음과 같은 방법과 경우를 제어하는 안전한 분산 훈련 루프를 제공합니다.
- 데이터 로드
- 예외 처리
- 체크포인트 파일 생성 및 장애 복구
- TensorBoard에 대한 요약 저장
Estimator로 애플리케이션을 작성할 때는 데이터 입력 파이프라인을 모델과 분리해야 합니다. 분리를 통해 서로 다른 데이터세트로 실험을 단순화합니다.
사전 작성 Estimator 사용하기
사전 작성 Estimator를 사용하면 기본 TensorFlow API보다 개념이 훨씬 높은 수준에서 작업할 수 있습니다. Estimator가 모든 "배관"을 처리하기 때문에 더 이상 계산 그래프 또는 세션 생성에 대해 걱정할 필요가 없습니다. 또한, 사전 작성 Estimator를 사용하면 최소한의 코드 변경만으로 다양한 모델 아키텍처를 실험할 수 있습니다. 예를 들어, tf.estimator.DNNClassifier는 밀집된 피드 전달 신경망을 기반으로 분류 모델을 훈련하는 사전 작성 Estimator 클래스입니다.
사전 작성된 Estimator에 의존하는 TensorFlow 프로그램은 일반적으로 다음 네 단계로 구성됩니다.
1. 입력 함수를 작성합니다.
예를 들어, 훈련 세트를 가져오는 함수 하나와 테스트 세트를 가져오는 다른 함수 하나를 만들 수 있습니다. Estimator는 입력이 한 쌍의 객체로 형식화될 것으로 예상합니다.
- 키가 특성 이름이고 값이 해당 특성 데이터를 포함하는 Tensor(또는 SparseTensors)인 사전
- 하나 이상의 레이블을 포함하는 Tensor
input_fn은 해당 형식의 쌍을 생성하는 tf.data.Dataset을 반환해야 합니다.
예를 들어, 다음 코드는 Titanic 데이터세트의 train.csv 파일에서 tf.data.Dataset를 빌드합니다.
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_fn은 tf.Graph에서 그래픽 텐서를 포함하는 (features_dics, labels) 쌍을 직접 반환할 수 있지만, 상수 반환과 같은 간단한 경우를 제외하고 오류가 발생하기 쉽습니다.
2. 특성 열을 정의합니다.
각 tf.feature_column은 특성 이름, 유형 및 모든 입력 전처리를 식별합니다.
예를 들어, 다음 조각은 3개의 특성 열을 만듭니다.
- 첫 번째는
age특성을 부동 소수점 입력으로 직접 사용합니다. - 두 번째는
class특성을 범주 입력으로 사용합니다. - 세 번째는
embark_town을 범주 입력으로 사용하지만,hashing trick을 사용하여 옵션을 열거할 필요가 없고 옵션 수를 설정합니다.
자세한 내용은 기능 열 튜토리얼을 확인하세요.
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)
3. 관련 사전 작성 Estimator를 인스턴스화합니다.
예를 들어, LinearClassifier라는 사전 작성된 Estimator의 샘플 인스턴스화는 다음과 같습니다.
model_dir = tempfile.mkdtemp()
model = tf.estimator.LinearClassifier(
model_dir=model_dir,
feature_columns=[embark, cls, age],
n_classes=2
)
INFO:tensorflow:Using default config.
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmp27g042fw', '_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}
자세한 내용은 선형 분류자 튜토리얼을 참조하세요.
4. 훈련, 평가 또는 추론 메서드를 호출합니다.
모든 Estimator는 train, evaluate 및 predict 메서드를 제공합니다.
model = model.train(input_fn=train_input_fn, steps=100)
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/training/training_util.py:396: Variable.initialized_value (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version. Instructions for updating: Use Variable.read_value. Variables in 2.X are initialized automatically both in eager and graph (inside tf.defun) contexts. 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/optimizers/optimizer_v2/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 INFO:tensorflow:Done calling model_fn. INFO:tensorflow:Create CheckpointSaverHook. 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/tmp27g042fw/model.ckpt. INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0... 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/tmp27g042fw/model.ckpt. INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100... INFO:tensorflow:Loss for final step: 0.66404176. 2022-12-14 21:57:27.516513: W tensorflow/core/kernels/data/cache_dataset_ops.cc:856] 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 2022-12-14T21:57:28 INFO:tensorflow:Graph was finalized. INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmp27g042fw/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 : 0.67038s INFO:tensorflow:Finished evaluation at 2022-12-14-21:57:29 INFO:tensorflow:Saving dict for global step 100: accuracy = 0.671875, accuracy_baseline = 0.65, auc = 0.6771763, auc_precision_recall = 0.5140185, average_loss = 0.6090718, global_step = 100, label/mean = 0.35, loss = 0.6090718, precision = 0.5443038, prediction/mean = 0.377406, recall = 0.38392857 INFO:tensorflow:Saving 'checkpoint_path' summary for global step 100: /tmpfs/tmp/tmp27g042fw/model.ckpt-100 accuracy : 0.671875 accuracy_baseline : 0.65 auc : 0.6771763 auc_precision_recall : 0.5140185 average_loss : 0.6090718 label/mean : 0.35 loss : 0.6090718 precision : 0.5443038 prediction/mean : 0.377406 recall : 0.38392857 global_step : 100 2022-12-14 21:57:29.065254: W tensorflow/core/kernels/data/cache_dataset_ops.cc:856] 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. INFO:tensorflow:Done calling model_fn. INFO:tensorflow:Graph was finalized. INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmp27g042fw/model.ckpt-100 INFO:tensorflow:Running local_init_op. INFO:tensorflow:Done running local_init_op. logits : [-1.3147551] logistic : [0.21169223] probabilities : [0.7883077 0.21169223] class_ids : [0] classes : [b'0'] all_class_ids : [0 1] all_classes : [b'0' b'1'] 2022-12-14 21:57:29.896072: W tensorflow/core/kernels/data/cache_dataset_ops.cc:856] 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.
사전 작성 Estimator의 이점
사전 작성 Estimator는 모범 사례를 인코딩하여 다음과 같은 이점을 제공합니다.
- 계산 그래프의 서로 다른 부분이 실행되어야 하는 위치를 결정하고 단일 머신 또는 클러스터에서 전략을 구현하는 모범 사례입니다.
- 이벤트(요약) 작성 및 보편적으로 유용한 요약에 대한 모범 사례입니다.
사전 작성 Estimator를 사용하지 않는 경우, 앞의 기능을 직접 구현해야 합니다.
사용자 정의 Estimator
모든 Estimator의 핵심은(사전 작성이든 사용자 정의이든) 훈련, 평가 및 예측을 빌드하는 메서드인 모델 함수, model_fn입니다. 사전 작성된 Estimator를 사용하는 경우, 다른 사람이 이미 모델 함수를 구현했습니다. 사용자 정의 Estimator에 의존할 때는 모델 함수를 직접 작성해야 합니다.
참고: 사용자 정의
model_fn은 여전히 1.x 스타일 그래프 모드에서 실행됩니다. 이는 즉시 실행이 없고 자동 제어 종속성이 없음을 의미합니다. 사용자 정의model_fn을 사용하여tf.estimator에서 마이그레이션할 계획을 세워야 합니다. 대체 API는tf.keras및tf.distribute입니다. 훈련의 일부에Estimator가 여전히 필요한 경우,tf.keras.estimator.model_to_estimator변환기를 사용하여keras.Model에서Estimator를 생성할 수 있습니다.
Keras 모델에서 Estimator 생성하기
tf.keras.estimator.model_to_estimator를 사용하여 기존 Keras 모델을 Estimator로 변환할 수 있습니다. 이는 모델 코드를 현대화하려는 경우 유용하지만, 훈련 파이프라인에는 여전히 Estimator가 필요합니다.
Keras MobileNet V2 모델을 인스턴스화하고 옵티마이저, 손실 및 메트릭으로 모델을 컴파일하여 다음으로 훈련합니다.
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
컴파일된 Keras 모델에서 Estimator를 작성합니다. Keras 모델의 초기 모델 상태는 작성된 Estimator에서 유지됩니다.
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/tmprfi9191i
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/backend.py:451: 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(
2022-12-14 21:57:34.516781: W tensorflow/c/c_api.cc:291] Operation '{name:'block_14_depthwise_BN/moving_mean/Assign' id:1876 op device:{requested: '', assigned: ''} def:{ { {node block_14_depthwise_BN/moving_mean/Assign} } = AssignVariableOp[_has_manual_control_dependencies=true, dtype=DT_FLOAT, validate_shape=false](block_14_depthwise_BN/moving_mean, block_14_depthwise_BN/moving_mean/Initializer/zeros)} }' 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/tmprfi9191i', '_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/tmprfi9191i', '_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}
파생된 Estimator를 다른 Estimator와 마찬가지로 취급합니다.
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
훈련하려면, Estimator의 훈련 함수를 호출합니다.
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/tmprfi9191i/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/tmprfi9191i/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/tmprfi9191i/keras/keras_model.ckpt
INFO:tensorflow:Warm-starting from: /tmpfs/tmp/tmprfi9191i/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/tmprfi9191i/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmprfi9191i/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:loss = 0.6898192, step = 0
INFO:tensorflow:loss = 0.6898192, 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/tmprfi9191i/model.ckpt.
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmprfi9191i/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: 0.67396164.
INFO:tensorflow:Loss for final step: 0.67396164.
<tensorflow_estimator.python.estimator.estimator.EstimatorV2 at 0x7f4ca8193a00>
마찬가지로, 평가하려면 Estimator의 평가 함수를 호출합니다.
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/engine/training_v1.py:2333: 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 2022-12-14T21:57:48
INFO:tensorflow:Starting evaluation at 2022-12-14T21:57:48
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmprfi9191i/model.ckpt-50
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmprfi9191i/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 : 2.30846s
INFO:tensorflow:Inference Time : 2.30846s
INFO:tensorflow:Finished evaluation at 2022-12-14-21:57:50
INFO:tensorflow:Finished evaluation at 2022-12-14-21:57:50
INFO:tensorflow:Saving dict for global step 50: accuracy = 0.515625, global_step = 50, loss = 0.66195524
INFO:tensorflow:Saving dict for global step 50: accuracy = 0.515625, global_step = 50, loss = 0.66195524
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50: /tmpfs/tmp/tmprfi9191i/model.ckpt-50
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50: /tmpfs/tmp/tmprfi9191i/model.ckpt-50
{'accuracy': 0.515625, 'loss': 0.66195524, 'global_step': 50}
자세한 내용은 tf.keras.estimator.model_to_estimator의 설명서를 참조하세요.
Estimator로 객체 기반 체크포인트 저장하기
Estimator는 기본적으로 체크포인트 가이드에 설명된 객체 그래프가 아닌 변수 이름으로 체크포인트를 저장합니다. tf.train.Checkpoint는 이름 기반 체크포인트를 읽지만, 모델의 일부를 Estimator의 model_fn 외부로 이동할 때 변수 이름이 변경될 수 있습니다. 포워드 호환성을 위해 객체 기반 체크포인트를 저장하면 Estimator 내부에서 모델을 훈련한 다음 모델 외부에서 사용하기가 더 쉽습니다.
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 = 4.6036353, step = 0
INFO:tensorflow:loss = 4.6036353, 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 ./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: 46.2618.
INFO:tensorflow:Loss for final step: 46.2618.
<tensorflow_estimator.python.estimator.estimator.EstimatorV2 at 0x7f4d384cb4f0>
tf.train.Checkpoint는 model_dir에서 Estimator의 체크포인트를 로드할 수 있습니다.
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
Estimator에서 SavedModel 내보내기
Estimator는 tf.Estimator.export_saved_model를 통해 SavedModels를 내보냅니다.
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/tmp5fjdzy_s
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmp5fjdzy_s
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmp5fjdzy_s', '_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/tmp5fjdzy_s', '_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/tmp5fjdzy_s/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmp5fjdzy_s/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/tmp5fjdzy_s/model.ckpt.
INFO:tensorflow:Saving checkpoints for 50 into /tmpfs/tmp/tmp5fjdzy_s/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: 0.46440706.
INFO:tensorflow:Loss for final step: 0.46440706.
<tensorflow_estimator.python.estimator.canned.linear.LinearClassifierV2 at 0x7f4c3425d970>
Estimator를 저장하려면 serving_input_receiver를 만들어야 합니다. 이 함수는 SavedModel에서 수신한 원시 데이터를 구문 분석하는 tf.Graph의 일부를 빌드합니다.
tf.estimator.export 모듈에는 이러한 receivers를 빌드하는 데 도움이 되는 함수가 포함되어 있습니다.
다음 코드는 tf-serving과 함께 자주 사용되는 직렬화된 tf.Example 프로토콜 버퍼를 허용하는 feature_columns을 기반으로 receiver를 빌드합니다.
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)
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:146: 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://www.tensorflow.org/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:146: 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://www.tensorflow.org/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:84: 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://www.tensorflow.org/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:84: 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://www.tensorflow.org/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/tmp5fjdzy_s/model.ckpt-50 INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmp5fjdzy_s/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/tmp8wjmhouf/from_estimator/temp-1671055072/saved_model.pb INFO:tensorflow:SavedModel written to: /tmpfs/tmp/tmp8wjmhouf/from_estimator/temp-1671055072/saved_model.pb
Python에서 해당 모델을 로드하고 실행할 수도 있습니다.
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([[0.58565766]], dtype=float32)>, 'class_ids': <tf.Tensor: shape=(1, 1), dtype=int64, numpy=array([[1]])>, 'all_classes': <tf.Tensor: shape=(1, 2), dtype=string, numpy=array([[b'0', b'1']], dtype=object)>, 'logits': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.34604275]], dtype=float32)>, 'classes': <tf.Tensor: shape=(1, 1), dtype=string, numpy=array([[b'1']], dtype=object)>, 'probabilities': <tf.Tensor: shape=(1, 2), dtype=float32, numpy=array([[0.41434237, 0.58565766]], dtype=float32)>, 'all_class_ids': <tf.Tensor: shape=(1, 2), dtype=int32, numpy=array([[0, 1]], dtype=int32)>}
{'logistic': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.22115663]], dtype=float32)>, 'class_ids': <tf.Tensor: shape=(1, 1), dtype=int64, numpy=array([[0]])>, 'all_classes': <tf.Tensor: shape=(1, 2), dtype=string, numpy=array([[b'0', b'1']], dtype=object)>, 'logits': <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[-1.2589388]], dtype=float32)>, 'classes': <tf.Tensor: shape=(1, 1), dtype=string, numpy=array([[b'0']], dtype=object)>, 'probabilities': <tf.Tensor: shape=(1, 2), dtype=float32, numpy=array([[0.77884334, 0.22115664]], dtype=float32)>, 'all_class_ids': <tf.Tensor: shape=(1, 2), dtype=int32, numpy=array([[0, 1]], dtype=int32)>}
tf.estimator.export.build_raw_serving_input_receiver_fn를 사용하여 tf.train.Example이 아닌 원시 텐서를 사용하는 입력 함수를 만들 수 있습니다.
Estimator와 함께 tf.distribute.Strategy 사용하기(제한적 지원)
tf.estimator는 원래 비동기 매개변수 서버 접근 방식을 지원했던 분산 훈련 TensorFlow API입니다. tf.estimator는 이제 tf.distribute.Strategy를 지원합니다. tf.estimator를 사용하는 경우, 코드를 거의 변경하지 않고 분산 훈련으로 변경할 수 있습니다. 이를 통해 Estimator 사용자는 이제 여러 GPU와 여러 작업자에서 동기식 분산 훈련을 수행하고 TPU도 사용할 수 있습니다. 그러나 Estimator의 이러한 지원은 제한적입니다. 자세한 내용은 현재 지원되는 기능 섹션을 참조하세요.
Estimator에서 tf.distribute.Strategy의 사용은 Keras의 경우와는 약간 다릅니다. 이제 strategy.scope를 사용하는 대신 Estimator용 RunConfig에 전략 객체를 전달합니다.
자세한 내용은 배포 교육 가이드를 참조하세요.
다음은 사전 작성 Estimator LinearRegressor 및 MirroredStrategy로 그 방법을 보여주는 코드 조각입니다.
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:Using temporary folder as model directory: /tmpfs/tmp/tmpvi5vs2vt
WARNING:tensorflow:Using temporary folder as model directory: /tmpfs/tmp/tmpvi5vs2vt
INFO:tensorflow:Using config: {'_model_dir': '/tmpfs/tmp/tmpvi5vs2vt', '_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 0x7f4c143873d0>, '_device_fn': None, '_protocol': None, '_eval_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f4c143873d0>, '_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/tmpvi5vs2vt', '_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 0x7f4c143873d0>, '_device_fn': None, '_protocol': None, '_eval_distribute': <tensorflow.python.distribute.mirrored_strategy.MirroredStrategy object at 0x7f4c143873d0>, '_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}
여기에서는 사전 작성된 Estimator를 사용하지만 사용자 정의 Estimator에서도 같은 코드가 동작합니다. train_distribute는 훈련이 배포되는 방식을 결정하고, eval_distribute는 평가가 배포되는 방식을 결정합니다. 이것은 훈련과 평가 모두에 같은 전략을 사용하는 Keras와의 또 다른 차이점입니다.
이제 입력 함수로 Estimator를 훈련하고 평가할 수 있습니다.
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:1244: 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:1244: 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:461: 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.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.
Instructions for updating:
Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.9/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.
Instructions for updating:
Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089
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:batch_all_reduce: 2 all-reduces with algorithm = nccl, num_packs = 1
INFO:tensorflow:batch_all_reduce: 2 all-reduces with algorithm = 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/tmpvi5vs2vt/model.ckpt.
INFO:tensorflow:Saving checkpoints for 0 into /tmpfs/tmp/tmpvi5vs2vt/model.ckpt.
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...
2022-12-14 21:57:58.395590: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:57:58.396844: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:57:58.401007: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:57:58.401454: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:57:58.406655: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:57:58.407105: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:57:58.424768: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:57:58.425212: 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/tmpvi5vs2vt/model.ckpt.
INFO:tensorflow:Saving checkpoints for 10 into /tmpfs/tmp/tmpvi5vs2vt/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 2022-12-14T21:58:01
INFO:tensorflow:Starting evaluation at 2022-12-14T21:58:01
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpvi5vs2vt/model.ckpt-10
INFO:tensorflow:Restoring parameters from /tmpfs/tmp/tmpvi5vs2vt/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.
2022-12-14 21:58:01.363673: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:58:01.364965: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:58:01.367936: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:58:01.368382: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:58:01.378875: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:58:01.379318: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorGetNextFromShard' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorGetNextFromShard} }
. Registered: device='CPU'
2022-12-14 21:58:01.385748: W tensorflow/core/grappler/utils/graph_view.cc:836] No registered 'MultiDeviceIteratorFromStringHandle' OpKernel for GPU devices compatible with node { {node MultiDeviceIteratorFromStringHandle} }
. Registered: device='CPU'
2022-12-14 21:58:01.386209: 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 : 0.67539s
INFO:tensorflow:Inference Time : 0.67539s
INFO:tensorflow:Finished evaluation at 2022-12-14-21:58:01
INFO:tensorflow:Finished evaluation at 2022-12-14-21:58:01
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/tmpvi5vs2vt/model.ckpt-10
INFO:tensorflow:Saving 'checkpoint_path' summary for global step 10: /tmpfs/tmp/tmpvi5vs2vt/model.ckpt-10
{'average_loss': 1.4210855e-14,
'label/mean': 1.0,
'loss': 1.4210855e-14,
'prediction/mean': 0.99999994,
'global_step': 10}
여기서 주목할 Estimator와 Keras의 또 다른 차이점은 입력 처리입니다. Keras에서는 데이터세트의 각 배치가 여러 복제본에 자동으로 분할됩니다. 그러나 Estimator에서는 배치의 자동 분할을 수행하지 않으며 여러 작업자 간에 데이터를 자동으로 샤딩하지도 않습니다. 작업자 및 기기 간에 데이터를 배포하는 방식을 자신이 완전히 제어할 수 있으며 데이터의 배포 방법을 지정하려면 input_fn을 제공해야 합니다.
input_fn은 작업자당 한 번 호출되므로 작업자당 하나의 데이터세트를 제공합니다. 그런 다음 해당 데이터세트의 배치 하나가 해당 작업자의 복제본 하나에 공급되어 작업자 1개에서 복제본 N개에 대해 N개의 배치를 소비합니다. 즉, input_fn에서 반환된 데이터세트는 크기 PER_REPLICA_BATCH_SIZE의 배치를 제공해야 합니다. 그리고 단계의 전역 배치 크기는 PER_REPLICA_BATCH_SIZE * strategy.num_replicas_in_sync로 얻을 수 있습니다.
다중 작업자 훈련을 수행할 때는 작업자 간에 데이터를 분할하거나 각 작업자에 대해 임의의 시드로 셔플해야 합니다. Estimator를 사용한 다중 작업자 훈련 튜토리얼에서 이를 수행하는 방법에 대한 예를 볼 수 있습니다.
마찬가지로 다중 작업자 및 매개변수 서버 전략도 사용할 수 있습니다. 코드는 동일하게 유지되지만, tf.estimator.train_and_evaluate를 사용하고 클러스터에서 실행 중인 각 바이너리에 대해 TF_CONFIG 환경 변수를 설정해야 합니다.
현재 지원되는 것은 무엇입니까?
TPUStrategy를 제외한 모든 전략을 사용하여 Estimator로 훈련하는 작업은 제한적으로 지원됩니다. 기본적인 훈련과 평가는 동작하지만, v1.train.Scaffold와 같은 여러 고급 기능은 동작하지 않습니다. 또한, 이 통합에는 다양한 버그가 있을 수도 있으며 현재로서는 이 지원을 적극적으로 개선할 계획이 없습니다(대신 Keras와 사용자 정의 훈련 루프의 지원에 중점을 둠). 가능하다면 해당 API에 tf.distribute를 대신 사용하는 것이 좋습니다.
| 훈련 API | MirroredStrategy | TPUStrategy | MultiWorkerMirroredStrategy | CentralStorageStrategy | ParameterServerStrategy |
|---|---|---|---|---|---|
| Estimator API | 제한적인 지원 | 지원되지 않음 | 제한적인 지원 | 제한적인 지원 | 제한적인 지원 |
예제 및 튜토리얼
다음은 Estimator로 다양한 전략을 사용하는 방법을 보여주는 몇 가지 엔드 투 엔드 예제입니다.
- Estimator를 사용한 다중 작업자 훈련 튜토리얼에서는 MNIST 데이터세트에서
MultiWorkerMirroredStrategy를 사용하여 여러 작업자로 교육하는 방법을 보여줍니다. - Kubernetes 템플릿을 사용하여
tensorflow/ecosystem에서 배포 전략으로 다중 작업자 훈련을 실행하는 엔드 투 엔드 예제. 이 예제는 Keras 모델로 시작하며tf.keras.estimator.model_to_estimatorAPI를 사용하여 이를 Estimator로 변환합니다. MirroredStrategy또는MultiWorkerMirroredStrategy를 사용하여 훈련할 수 있는 공식 ResNet50 모델