tensorflow的slim包集成了很多tensorflow中的高效函数，也其实就是原版tensorflow的瘦身(slim) API1.4
1. 导入
1 import tensorflow.contrib.slim as slim
2. 定义
1 # Copyright 2016 The TensorFlow Authors. All Rights Reserved. 2 # 3 # Licensed under the Apache License, Version 2.0 (the "License"); 4 # you may not use this file except in compliance with the License. 5 # You may obtain a copy of the License at 6 # 7 # http://www.apache.org/licenses/LICENSE-2.0 8 # 9 # Unless required by applicable law or agreed to in writing, software 10 # distributed under the License is distributed on an "AS IS" BASIS, 11 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 # See the License for the specific language governing permissions and 13 # limitations under the License. 14 # ============================================================================== 15 """Slim is an interface to contrib functions, examples and models. 16 17 TODO(nsilberman): flesh out documentation. 18 """ 19 20 from __future__ import absolute_import 21 from __future__ import division 22 from __future__ import print_function 23 24 # pylint: disable=unused-import,line-too-long,g-importing-member,wildcard-import 25 # TODO(jart): Delete non-slim imports 26 from tensorflow.contrib import losses 27 from tensorflow.contrib import metrics 28 from tensorflow.contrib.framework.python.ops.arg_scope import * 29 from tensorflow.contrib.framework.python.ops.variables import * 30 from tensorflow.contrib.layers.python.layers import * 31 from tensorflow.contrib.layers.python.layers.initializers import * 32 from tensorflow.contrib.layers.python.layers.regularizers import * 33 from tensorflow.contrib.slim.python.slim import evaluation 34 from tensorflow.contrib.slim.python.slim import learning 35 from tensorflow.contrib.slim.python.slim import model_analyzer 36 from tensorflow.contrib.slim.python.slim import queues 37 from tensorflow.contrib.slim.python.slim import summaries 38 from tensorflow.contrib.slim.python.slim.data import data_decoder 39 from tensorflow.contrib.slim.python.slim.data import data_provider 40 from tensorflow.contrib.slim.python.slim.data import dataset 41 from tensorflow.contrib.slim.python.slim.data import dataset_data_provider 42 from tensorflow.contrib.slim.python.slim.data import parallel_reader 43 from tensorflow.contrib.slim.python.slim.data import prefetch_queue 44 from tensorflow.contrib.slim.python.slim.data import tfexample_decoder 45 from tensorflow.python.util.all_util import make_all 46 # pylint: enable=unused-import,line-too-long,g-importing-member,wildcard-import 47 48 __all__ = make_all(__name__)
只是其他函数的一个简化引用，举例：
1 from __future__ import absolute_import 2 from __future__ import division 3 from __future__ import print_function 4 5 # pylint: disable=unused-import,wildcard-import 6 from tensorflow.contrib.framework.python.framework import * 7 from tensorflow.contrib.framework.python.ops import * 8 # pylint: enable=unused-import,wildcard-import 9 10 from tensorflow.python.framework.ops import prepend_name_scope 11 from tensorflow.python.framework.ops import strip_name_scope 12 13 from tensorflow.python.util.all_util import remove_undocumented 14 15 _allowed_symbols = ['nest'] 16 17 remove_undocumented(__name__, allowed_exception_list=_allowed_symbols)
1 import tf.contrib.framework.arg_scope as arg_scope 2 import tf.contrib.slim.arg_scope as arg_scope
上面两种导入方式其实是对同一个文件的不同导入方式
3. slim的简化使用例子，和原版对比
https://www.jianshu.com/p/e2ada4ddae9a
4.
slim中函数汇总如下：
1 import tensorflow as tf
2 import tf.contrib.slim as slim
1). slim.losses
1 __all__ = ["absolute_difference", @deprecated("2016-12-30", "Use tf.losses.absolute_difference instead.") 2 "add_loss", @deprecated("2016-12-30", "Use tf.losses.add_loss instead.") 3 "cosine_distance", @deprecated("2016-12-30", "Use tf.losses.cosine_distance instead.") 4 "compute_weighted_loss", @deprecated("2016-12-30", "Use tf.losses.compute_weighted_loss instead.") 5 "get_losses", @deprecated("2016-12-30", "Use tf.losses.get_losses instead.") 6 "get_regularization_losses", @deprecated("2016-12-30", "Use tf.losses.get_regularization_losses instead.") 7 "get_total_loss", @deprecated("2016-12-30", "Use tf.losses.get_total_loss instead.") 8 "hinge_loss", @deprecated("2016-12-30","Use tf.losses.hinge_loss instead. Note that the order of the logits and labels arguments has been changed, and to stay unweighted, reduction=Reduction.NONE") 9 "log_loss", @deprecated("2016-12-30",Use tf.losses.log_loss instead. Note that the order of the predictions and labels arguments has been changed.") 10 "mean_pairwise_squared_error", @deprecated("2016-12-30","Use tf.losses.mean_pairwise_squared_error instead. Note that the order of the predictions and labels arguments has been changed.") 11 "mean_squared_error", @deprecated("2016-12-30", "Use tf.losses.mean_squared_error instead.") 12 "sigmoid_cross_entropy", @deprecated("2016-12-30","Use tf.losses.sigmoid_cross_entropy instead. Note that the order of the predictions and labels arguments has been changed.") 13 "softmax_cross_entropy", @deprecated("2016-12-30","Use tf.losses.softmax_cross_entropy instead. Note that the order of the logits and labels arguments has been changed.") 14 "sparse_softmax_cross_entropy"] @deprecated("2016-12-30","Use tf.losses.sparse_softmax_cross_entropy instead. Note that the order of the logits and labels arguments has been changed.")
就是说slim基本上已经不支持losses操作了，全部改成 tf.losses的操作
为什么？ 可能是想使分类更加具有层次吧，loss就该处理loss, 而不是有什么cosine_distance
1 @@add_loss 2 @@get_losses 3 @@get_regularization_loss 4 @@get_regularization_losses 5 @@get_total_loss
其他的tf.losses函数在另外两个文件中有定义
2） slim.metrics
1 @@streaming_accuracy 2 @@streaming_mean 3 @@streaming_recall 4 @@streaming_recall_at_thresholds 5 @@streaming_precision 6 @@streaming_precision_at_thresholds 7 @@streaming_auc 8 @@streaming_curve_points 9 @@streaming_recall_at_k 10 @@streaming_mean_absolute_error 11 @@streaming_mean_iou 12 @@streaming_mean_relative_error 13 @@streaming_mean_squared_error 14 @@streaming_mean_tensor 15 @@streaming_root_mean_squared_error 16 @@streaming_covariance 17 @@streaming_pearson_correlation 18 @@streaming_mean_cosine_distance 19 @@streaming_percentage_less 20 @@streaming_sensitivity_at_specificity 21 @@streaming_sparse_average_precision_at_k 22 @@streaming_sparse_average_precision_at_top_k 23 @@streaming_sparse_precision_at_k 24 @@streaming_sparse_precision_at_top_k 25 @@streaming_sparse_recall_at_k 26 @@streaming_specificity_at_sensitivity 27 @@streaming_concat 28 @@streaming_false_negatives 29 @@streaming_false_negatives_at_thresholds 30 @@streaming_false_positives 31 @@streaming_false_positives_at_thresholds 32 @@streaming_true_negatives 33 @@streaming_true_negatives_at_thresholds 34 @@streaming_true_positives 35 @@streaming_true_positives_at_thresholds 36 @@sparse_recall_at_top_k 37 @@auc_using_histogram 38 @@accuracy 39 @@aggregate_metrics 40 @@aggregate_metric_map 41 @@confusion_matrix 42 @@set_difference 43 @@set_intersection 44 @@set_size 45 @@set_union
3) slim.arg_scope
1 __all__ = ['arg_scope', 2 'add_arg_scope', 3 'has_arg_scope', 4 'arg_scoped_arguments']
4) slim.variables
1 __all__ = ['add_model_variable', 2 'assert_global_step', @deprecated(None, "Please switch to tf.train.assert_global_step") 3 'assert_or_get_global_step', 4 'assign_from_checkpoint', 5 'assign_from_checkpoint_fn', 6 'assign_from_values', 7 'assign_from_values_fn', 8 'create_global_step', @deprecated(None, "Please switch to tf.train.create_global_step") 9 'filter_variables', 10 'get_global_step', @deprecated(None, "Please switch to tf.train.get_global_step") 11 'get_or_create_global_step', @deprecated(None, "Please switch to tf.train.get_or_create_global_step") 12 'get_local_variables', 13 'get_model_variables', 14 'get_trainable_variables', 15 'get_unique_variable', 16 'get_variables_by_name', 17 'get_variables_by_suffix', 18 'get_variable_full_name', 19 'get_variables_to_restore', 20 'get_variables', 21 'local_variable', 22 'model_variable', 23 'variable', 24 'VariableDeviceChooser', 25 'zero_initializer']
主要是把slim.variables中关于global_step的函数全部移除了，改为tf.train中global_step的函数
5） from tensorflow.contrib.layers.python.layers import *
1 from tensorflow.contrib.layers.python.layers.embedding_ops import * 2 from tensorflow.contrib.layers.python.layers.encoders import * 3 from tensorflow.contrib.layers.python.layers.feature_column import * 4 from tensorflow.contrib.layers.python.layers.feature_column_ops import * 5 from tensorflow.contrib.layers.python.layers.initializers import * 6 from tensorflow.contrib.layers.python.layers.layers import * 7 from tensorflow.contrib.layers.python.layers.normalization import * 8 from tensorflow.contrib.layers.python.layers.optimizers import * 9 from tensorflow.contrib.layers.python.layers.regularizers import * 10 from tensorflow.contrib.layers.python.layers.summaries import * 11 from tensorflow.contrib.layers.python.layers.target_column import * 12 from tensorflow.contrib.layers.python.ops.bucketization_op import * 13 from tensorflow.contrib.layers.python.ops.sparse_feature_cross_op import *
embedding_ops.py
1 __all__ = [ 2 "safe_embedding_lookup_sparse", "scattered_embedding_lookup", 3 "scattered_embedding_lookup_sparse", "embedding_lookup_unique", 4 "embedding_lookup_sparse_with_distributed_aggregation" 5 ]
encoder.py
1__all__ = ['bow_encoder', 'embed_sequence']
initializer.py
1 __all__ = ['xavier_initializer', 'xavier_initializer_conv2d', 2 'variance_scaling_initializer']
layers.py
1 __all__ = ['avg_pool2d', 2 'avg_pool3d', 3 'batch_norm', 4 'bias_add', 5 'conv2d', 6 'conv3d', 7 'conv2d_in_plane', 8 'conv2d_transpose', 9 'conv3d_transpose', 10 'convolution', 11 'convolution2d', 12 'convolution2d_in_plane', 13 'convolution2d_transpose', 14 'convolution3d', 15 'convolution3d_transpose', 16 'dropout', 17 'elu', 18 'flatten', 19 'fully_connected', 20 'GDN', 21 'gdn', 22 'layer_norm', 23 'linear', 24 'pool', 25 'max_pool2d', 26 'max_pool3d', 27 'one_hot_encoding', 28 'relu', 29 'relu6', 30 'repeat', 31 'scale_gradient', 32 'separable_conv2d', 33 'separable_convolution2d', 34 'softmax', 35 'spatial_softmax', 36 'stack', 37 'unit_norm', 38 'legacy_fully_connected', 39 'legacy_linear', 40 'legacy_relu', 41 'maxout']
normalization.py
1 __all__ = [ 2 'instance_norm', 3 ]
optimizer.py
1 __all__ = [ 2 'optimize_loss', 3 'adaptive_clipping_fn', 4 'gradient_clipping', 5 ]
regularizers.py
1 __all__ = ['l1_regularizer',
2 'l2_regularizer',
3 'l1_l2_regularizer',
4 'sum_regularizer',
5 'apply_regularization']
summaries.py
1 __all__ = [ 2 'summarize_tensor', 3 'summarize_activation', 4 'summarize_tensors', 5 'summarize_collection', 6 'summarize_variables', 7 'summarize_weights', 8 'summarize_biases', 9 'summarize_activations', 10 ]
target_column.py
1 @deprecated("2016-11-12", "This file will be removed after the deprecation date. Please switch to third_party/tensorflow/contrib/learn/python/learn/estimators/head.py")
utils.py
1 __all__ = ['collect_named_outputs', 2 'constant_value', 3 'static_cond', 4 'smart_cond', 5 'get_variable_collections', 6 'two_element_tuple', 7 'n_positive_integers', 8 'channel_dimension', 9 'last_dimension']
feature_column.py feature_column_ops.py
6)
slim.evaluation
1 __all__ = [ 2 'evaluate_once', 3 'evaluation_loop', 4 'wait_for_new_checkpoint', 5 'checkpoints_iterator', 6 ]
slim.learning
1 __all__ = [ 2 'add_gradients_summaries', 'clip_gradient_norms', 'multiply_gradients', 3 'create_train_op', 'train_step', 'train' 4 ]
slim.model_analyzer
1 __all__ = [ 2 'tensor_description', 3 'analyze_ops', 4 'analyze_vars', 5 ]
slim.queues
1 __all__ = [ 2 'NestedQueueRunnerError', 3 'QueueRunners', 4 ]
slim.summaries
1 __all__ = [ 2 'add_histogram_summary', 3 'add_image_summary', 4 'add_scalar_summary', 5 'add_zero_fraction_summary', 6 'add_histogram_summaries', 7 'add_image_summaries', 8 'add_scalar_summaries', 9 'add_zero_fraction_summaries', 10 ]
7)
slim.data_decoder
slim.data_provider
class DataProvider(object): def get(self,items): def list_items(self): def num_samples(self):
slim.dataset
1 class Dataset(object): 2 """Represents a Dataset specification.""" 3 4 def __init__(self, data_sources, reader, decoder, num_samples, 5 items_to_descriptions, **kwargs):
slim.dataset_data_provider
slim.parallel_reader
slim.prefetch_queue
slim.tfexample_decoder
8)
1 from tensorflow.contrib.slim.python.slim.nets import alexnet 2 from tensorflow.contrib.slim.python.slim.nets import inception 3 from tensorflow.contrib.slim.python.slim.nets import overfeat 4 from tensorflow.contrib.slim.python.slim.nets import resnet_utils 5 from tensorflow.contrib.slim.python.slim.nets import resnet_v1 6 from tensorflow.contrib.slim.python.slim.nets import resnet_v2 7 from tensorflow.contrib.slim.python.slim.nets import vgg 8 from tensorflow.python.util.all_util import make_all
1 alexnet.py 2 inception.py 3 inception_v1.py 4 inception_v2.py 5 inception_v3.py 6 overfeat.py 7 resnet_utils.py 8 resnet_v1.py 9 resnet_v2.py 10 vgg.py
9) slim.make_all
可以把所有文件中__all__的方法返回