1096 lines (1050 loc) · 79.5 KB

TensorFlow Python reference documentation

Building Graphs:
- add_to_collection
- as_dtype
- container
- control_dependencies
- convert_to_tensor
- convert_to_tensor_or_indexed_slices
- device
- DeviceSpec
- Dimension
- DType
- get_collection
- get_collection_ref
- get_default_graph
- get_seed
- Graph
- GraphKeys
- import_graph_def
- load_file_system_library
- load_op_library
- name_scope
- NoGradient
- NotDifferentiable
- op_scope
- Operation
- register_tensor_conversion_function
- RegisterGradient
- RegisterShape
- reset_default_graph
- Tensor
- TensorShape
Asserts and boolean checks.:
- assert_equal
- assert_greater
- assert_greater_equal
- assert_integer
- assert_less
- assert_less_equal
- assert_negative
- assert_non_negative
- assert_non_positive
- assert_positive
- assert_proper_iterable
- assert_rank
- assert_rank_at_least
- assert_type
- is_non_decreasing
- is_numeric_tensor
- is_strictly_increasing
Constants, Sequences, and Random Values:
- constant
- fill
- linspace
- multinomial
- ones
- ones_like
- random_crop
- random_gamma
- random_normal
- random_shuffle
- random_uniform
- range
- set_random_seed
- truncated_normal
- zeros
- zeros_like
Variables:
- all_variables
- assert_variables_initialized
- assign
- assign_add
- assign_sub
- constant_initializer
- count_up_to
- device
- export_meta_graph
- fixed_size_partitioner
- get_checkpoint_state
- get_variable
- get_variable_scope
- global_variables
- global_variables_initializer
- import_meta_graph
- IndexedSlices
- initialize_all_tables
- initialize_all_variables
- initialize_local_variables
- initialize_variables
- is_variable_initialized
- latest_checkpoint
- local_variables
- local_variables_initializer
- make_template
- min_max_variable_partitioner
- model_variables
- moving_average_variables
- no_regularizer
- ones_initializer
- orthogonal_initializer
- random_normal_initializer
- random_uniform_initializer
- report_uninitialized_variables
- Saver
- scatter_add
- scatter_div
- scatter_mul
- scatter_nd_add
- scatter_nd_sub
- scatter_nd_update
- scatter_sub
- scatter_update
- sparse_mask
- trainable_variables
- truncated_normal_initializer
- uniform_unit_scaling_initializer
- update_checkpoint_state
- Variable
- variable_axis_size_partitioner
- variable_op_scope
- variable_scope
- variables_initializer
- VariableScope
- zeros_initializer
Tensor Transformations:
- batch_to_space
- batch_to_space_nd
- bitcast
- boolean_mask
- cast
- concat
- concat_v2
- depth_to_space
- dequantize
- dynamic_partition
- dynamic_stitch
- expand_dims
- extract_image_patches
- fake_quant_with_min_max_args
- fake_quant_with_min_max_args_gradient
- fake_quant_with_min_max_vars
- fake_quant_with_min_max_vars_gradient
- fake_quant_with_min_max_vars_per_channel
- fake_quant_with_min_max_vars_per_channel_gradient
- gather
- gather_nd
- meshgrid
- one_hot
- pack
- pad
- quantize_v2
- quantized_concat
- rank
- required_space_to_batch_paddings
- reshape
- reverse
- reverse_sequence
- reverse_v2
- saturate_cast
- scatter_nd
- sequence_mask
- setdiff1d
- shape
- shape_n
- size
- slice
- space_to_batch
- space_to_batch_nd
- space_to_depth
- split
- split_v
- squeeze
- stack
- strided_slice
- string_to_number
- tile
- to_bfloat16
- to_double
- to_float
- to_int32
- to_int64
- transpose
- unique_with_counts
- unpack
- unstack
Math:
- abs
- accumulate_n
- acos
- add
- add_n
- argmax
- argmin
- asin
- atan
- batch_matmul
- betainc
- ceil
- cholesky
- cholesky_solve
- complex
- complex_abs
- conj
- cos
- count_nonzero
- cross
- cumprod
- cumsum
- diag
- diag_part
- digamma
- div
- divide
- edit_distance
- einsum
- erf
- erfc
- exp
- eye
- fft
- fft2d
- fft3d
- floor
- floor_div
- floordiv
- floormod
- ifft
- ifft2d
- ifft3d
- igamma
- igammac
- imag
- invert_permutation
- lbeta
- lgamma
- log
- log1p
- matmul
- matrix_band_part
- matrix_determinant
- matrix_diag
- matrix_diag_part
- matrix_inverse
- matrix_set_diag
- matrix_solve
- matrix_solve_ls
- matrix_transpose
- matrix_triangular_solve
- maximum
- minimum
- mod
- mul
- multiply
- neg
- negative
- polygamma
- pow
- real
- realdiv
- reciprocal
- reduce_all
- reduce_any
- reduce_logsumexp
- reduce_max
- reduce_mean
- reduce_min
- reduce_prod
- reduce_sum
- rint
- round
- rsqrt
- scalar_mul
- segment_max
- segment_mean
- segment_min
- segment_prod
- segment_sum
- self_adjoint_eig
- self_adjoint_eigvals
- setdiff1d
- sign
- sin
- sparse_segment_mean
- sparse_segment_sqrt_n
- sparse_segment_sum
- sqrt
- square
- squared_difference
- sub
- subtract
- svd
- tan
- trace
- transpose
- truediv
- truncatediv
- truncatemod
- unique
- unsorted_segment_sum
- where
- zeta
Strings:
- as_string
- decode_base64
- encode_base64
- reduce_join
- string_join
- string_split
- string_to_hash_bucket
- string_to_hash_bucket_fast
- string_to_hash_bucket_strong
- substr
Histograms:
- histogram_fixed_width
Control Flow:
- add_check_numerics_ops
- Assert
- case
- check_numerics
- cond
- count_up_to
- equal
- greater
- greater_equal
- group
- identity
- is_finite
- is_inf
- is_nan
- less
- less_equal
- logical_and
- logical_not
- logical_or
- logical_xor
- no_op
- not_equal
- Print
- select
- tuple
- verify_tensor_all_finite
- where
- while_loop
Higher Order Functions:
- foldl
- foldr
- map_fn
- scan
TensorArray Operations:
- concat
- gather
- pack
- split
- TensorArray
- unpack
Tensor Handle Operations:
- delete_session_tensor
- get_session_handle
- get_session_tensor
Images:
- adjust_brightness
- adjust_contrast
- adjust_gamma
- adjust_hue
- adjust_saturation
- central_crop
- convert_image_dtype
- crop_and_resize
- crop_to_bounding_box
- decode_gif
- decode_jpeg
- decode_png
- draw_bounding_boxes
- encode_jpeg
- encode_png
- extract_glimpse
- flip_left_right
- flip_up_down
- grayscale_to_rgb
- hsv_to_rgb
- non_max_suppression
- pad_to_bounding_box
- per_image_standardization
- random_brightness
- random_contrast
- random_flip_left_right
- random_flip_up_down
- random_hue
- random_saturation
- resize_area
- resize_bicubic
- resize_bilinear
- resize_image_with_crop_or_pad
- resize_images
- resize_nearest_neighbor
- rgb_to_grayscale
- rgb_to_hsv
- rot90
- sample_distorted_bounding_box
- transpose_image
Sparse Tensors:
- shape
- sparse_add
- sparse_concat
- sparse_fill_empty_rows
- sparse_maximum
- sparse_merge
- sparse_minimum
- sparse_reduce_sum
- sparse_reduce_sum_sparse
- sparse_reorder
- sparse_reset_shape
- sparse_reshape
- sparse_retain
- sparse_softmax
- sparse_split
- sparse_tensor_dense_matmul
- sparse_tensor_to_dense
- sparse_to_dense
- sparse_to_indicator
- sparse_transpose
- SparseTensor
- SparseTensorValue
Inputs and Readers:
- batch
- batch_join
- ConditionalAccumulator
- ConditionalAccumulatorBase
- decode_csv
- decode_json_example
- decode_raw
- FIFOQueue
- FixedLenFeature
- FixedLengthRecordReader
- FixedLenSequenceFeature
- IdentityReader
- input_producer
- limit_epochs
- match_filenames_once
- matching_files
- PaddingFIFOQueue
- parse_example
- parse_single_example
- parse_tensor
- placeholder
- placeholder_with_default
- PriorityQueue
- QueueBase
- RandomShuffleQueue
- range_input_producer
- read_file
- ReaderBase
- shuffle_batch
- shuffle_batch_join
- size
- slice_input_producer
- sparse_placeholder
- SparseConditionalAccumulator
- string_input_producer
- TextLineReader
- TFRecordReader
- VarLenFeature
- WholeFileReader
- write_file
Data IO (Python functions):
- tf_record_iterator
- TFRecordCompressionType
- TFRecordOptions
- TFRecordWriter
Neural Network:
- atrous_conv2d
- avg_pool
- avg_pool3d
- batch_normalization
- bias_add
- bidirectional_dynamic_rnn
- bidirectional_rnn
- compute_accidental_hits
- conv1d
- conv2d
- conv2d_transpose
- conv3d
- conv3d_transpose
- convolution
- crelu
- ctc_beam_search_decoder
- ctc_greedy_decoder
- ctc_loss
- depthwise_conv2d
- depthwise_conv2d_native
- dilation2d
- dropout
- dynamic_rnn
- elu
- embedding_lookup
- embedding_lookup_sparse
- erosion2d
- fixed_unigram_candidate_sampler
- fractional_avg_pool
- fractional_max_pool
- in_top_k
- l2_loss
- l2_normalize
- learned_unigram_candidate_sampler
- local_response_normalization
- log_poisson_loss
- log_softmax
- log_uniform_candidate_sampler
- max_pool
- max_pool3d
- max_pool_with_argmax
- moments
- nce_loss
- normalize_moments
- pool
- quantized_avg_pool
- quantized_max_pool
- quantized_relu_x
- raw_rnn
- relu
- relu6
- rnn
- sampled_softmax_loss
- separable_conv2d
- sigmoid
- sigmoid_cross_entropy_with_logits
- softmax
- softmax_cross_entropy_with_logits
- softplus
- softsign
- sparse_softmax_cross_entropy_with_logits
- state_saving_rnn
- sufficient_statistics
- tanh
- top_k
- uniform_candidate_sampler
- weighted_cross_entropy_with_logits
- weighted_moments
Neural Network RNN Cells:
- BasicLSTMCell
- BasicRNNCell
- DropoutWrapper
- EmbeddingWrapper
- GRUCell
- InputProjectionWrapper
- LSTMCell
- LSTMStateTuple
- MultiRNNCell
- OutputProjectionWrapper
- RNNCell
Running Graphs:
- AbortedError
- AlreadyExistsError
- CancelledError
- DataLossError
- DeadlineExceededError
- error_code_from_exception_type
- exception_type_from_error_code
- FailedPreconditionError
- get_default_session
- InteractiveSession
- InternalError
- InvalidArgumentError
- NotFoundError
- OpError
- OutOfRangeError
- PermissionDeniedError
- raise_exception_on_not_ok_status
- ResourceExhaustedError
- Session
- UnauthenticatedError
- UnavailableError
- UnimplementedError
- UnknownError
Training:
- AdadeltaOptimizer
- AdagradDAOptimizer
- AdagradOptimizer
- AdamOptimizer
- add_queue_runner
- AggregationMethod
- assert_global_step
- audio_summary
- basic_train_loop
- checkpoint_exists
- CheckpointSaverHook
- ChiefSessionCreator
- clip_by_average_norm
- clip_by_global_norm
- clip_by_norm
- clip_by_value
- ClusterSpec
- Coordinator
- do_quantize_training_on_graphdef
- exponential_decay
- ExponentialMovingAverage
- FtrlOptimizer
- generate_checkpoint_state_proto
- get_checkpoint_mtimes
- get_global_step
- global_norm
- global_step
- GradientDescentOptimizer
- gradients
- hessians
- histogram_summary
- image_summary
- inverse_time_decay
- LoggingTensorHook
- LooperThread
- merge_all_summaries
- merge_summary
- MomentumOptimizer
- MonitoredSession
- MonitoredTrainingSession
- NanLossDuringTrainingError
- NanTensorHook
- natural_exp_decay
- NewCheckpointReader
- Optimizer
- piecewise_constant
- polynomial_decay
- ProximalAdagradOptimizer
- ProximalGradientDescentOptimizer
- QueueRunner
- replica_device_setter
- RMSPropOptimizer
- Scaffold
- scalar_summary
- Server
- SessionCreator
- SessionManager
- SessionRunArgs
- SessionRunContext
- SessionRunHook
- SessionRunValues
- start_queue_runners
- StepCounterHook
- stop_gradient
- StopAtStepHook
- summary_iterator
- SummarySaverHook
- SummaryWriter
- SummaryWriterCache
- Supervisor
- SyncReplicasOptimizer
- SyncReplicasOptimizerV2
- WorkerSessionCreator
- write_graph
- zero_fraction
Wraps python functions:
- py_func
Summary Operations:
- audio
- FileWriter
- get_summary_description
- histogram
- image
- merge
- merge_all
- scalar
- SummaryDescription
- TaggedRunMetadata
- tensor_summary
Testing:
- assert_equal_graph_def
- Benchmark
- compute_gradient
- compute_gradient_error
- get_temp_dir
- is_built_with_cuda
- is_gpu_available
- main
- test_src_dir_path
- TestCase
BayesFlow Entropy (contrib):
- elbo_ratio
- entropy_shannon
- renyi_alpha
- renyi_ratio
BayesFlow Monte Carlo (contrib):
- expectation
- expectation_importance_sampler
- expectation_importance_sampler_logspace
BayesFlow Stochastic Graph (contrib):
- surrogate_loss
BayesFlow Stochastic Tensors (contrib):
- BaseStochasticTensor
- get_current_value_type
- MeanValue
- ObservedStochasticTensor
- SampleValue
- StochasticTensor
- value_type
BayesFlow Variational Inference (contrib):
- elbo
- elbo_with_log_joint
- ELBOForms
- register_prior
CRF (contrib):
- crf_binary_score
- crf_log_likelihood
- crf_log_norm
- crf_sequence_score
- crf_unary_score
- CrfForwardRnnCell
- viterbi_decode
Statistical distributions (contrib):
- Bernoulli
- BernoulliWithSigmoidP
- Beta
- BetaWithSoftplusAB
- Binomial
- Categorical
- Chi2
- Chi2WithAbsDf
- Dirichlet
- DirichletMultinomial
- Distribution
- Exponential
- ExponentialWithSoftplusLam
- Gamma
- GammaWithSoftplusAlphaBeta
- InverseGamma
- InverseGammaWithSoftplusAlphaBeta
- kl
- Laplace
- LaplaceWithSoftplusScale
- matrix_diag_transform
- Mixture
- Multinomial
- MultivariateNormalCholesky
- MultivariateNormalDiag
- MultivariateNormalDiagPlusVDVT
- MultivariateNormalDiagWithSoftplusStDev
- MultivariateNormalFull
- Normal
- normal_conjugates_known_sigma_posterior
- normal_conjugates_known_sigma_predictive
- NormalWithSoftplusSigma
- Poisson
- QuantizedDistribution
- RegisterKL
- StudentT
- StudentTWithAbsDfSoftplusSigma
- TransformedDistribution
- Uniform
- WishartCholesky
- WishartFull
Random variable transformations (contrib):
- Bijector
- Chain
- CholeskyOuterProduct
- Exp
- Identity
- Inline
- Invert
- ScaleAndShift
- SigmoidCentered
- SoftmaxCentered
- Softplus
FFmpeg (contrib):
- decode_audio
- encode_audio
Framework (contrib):
- add_arg_scope
- add_model_variable
- arg_scope
- arg_scoped_arguments
- assert_global_step
- assert_or_get_global_step
- assert_same_float_dtype
- assert_scalar_int
- assign_from_checkpoint
- assign_from_checkpoint_fn
- assign_from_values
- assign_from_values_fn
- convert_to_tensor_or_sparse_tensor
- create_global_step
- deprecated
- deprecated_arg_values
- deprecated_args
- get_global_step
- get_graph_from_inputs
- get_local_variables
- get_model_variables
- get_or_create_global_step
- get_unique_variable
- get_variables
- get_variables_by_name
- get_variables_by_suffix
- get_variables_to_restore
- has_arg_scope
- is_non_decreasing
- is_numeric_tensor
- is_strictly_increasing
- is_tensor
- local_variable
- model_variable
- reduce_sum_n
- variable
- VariableDeviceChooser
- with_same_shape
- with_shape
- zero_initializer
Graph Editor (contrib):
- add_control_inputs
- assign_renamed_collections_handler
- bypass
- check_cios
- compute_boundary_ts
- connect
- ControlOutputs
- copy
- copy_op_handler
- copy_with_input_replacements
- detach
- detach_control_inputs
- detach_control_outputs
- detach_inputs
- detach_outputs
- filter_ops
- filter_ops_from_regex
- filter_ts
- filter_ts_from_regex
- get_backward_walk_ops
- get_consuming_ops
- get_forward_walk_ops
- get_generating_ops
- get_name_scope_ops
- get_ops_ios
- get_tensors
- get_walks_intersection_ops
- get_walks_union_ops
- get_within_boundary_ops
- graph_replace
- keep_t_if_possible_handler
- make_list_of_op
- make_list_of_t
- make_placeholder_from_dtype_and_shape
- make_placeholder_from_tensor
- make_view
- make_view_from_scope
- matcher
- op_type
- OpMatcher
- ops
- ph
- placeholder_name
- remove_control_inputs
- replace_t_with_placeholder_handler
- reroute_a2b
- reroute_a2b_inputs
- reroute_a2b_outputs
- reroute_a2b_ts
- reroute_b2a
- reroute_b2a_inputs
- reroute_b2a_outputs
- reroute_b2a_ts
- select_ops
- select_ops_and_ts
- select_ts
- sgv
- sgv_scope
- SubGraphView
- swap
- swap_inputs
- swap_outputs
- swap_ts
- transform_op_if_inside_handler
- transform_op_in_place
- Transformer
- ts
Integrate (contrib):
- odeint
Layers (contrib):
- apply_regularization
- avg_pool2d
- batch_norm
- bucketized_column
- check_feature_columns
- convolution2d
- convolution2d_in_plane
- convolution2d_transpose
- create_feature_spec_for_parsing
- crossed_column
- embedding_column
- flatten
- fully_connected
- hashed_embedding_column
- input_from_feature_columns
- joint_weighted_sum_from_feature_columns
- l1_regularizer
- l2_regularizer
- layer_norm
- make_place_holder_tensors_for_base_features
- max_pool2d
- one_hot_column
- one_hot_encoding
- optimize_loss
- parse_feature_columns_from_examples
- parse_feature_columns_from_sequence_examples
- real_valued_column
- repeat
- safe_embedding_lookup_sparse
- separable_convolution2d
- shared_embedding_columns
- sparse_column_with_hash_bucket
- sparse_column_with_integerized_feature
- sparse_column_with_keys
- stack
- sum_regularizer
- summarize_activation
- summarize_activations
- summarize_collection
- summarize_tensor
- summarize_tensors
- unit_norm
- variance_scaling_initializer
- weighted_sparse_column
- weighted_sum_from_feature_columns
- xavier_initializer
- xavier_initializer_conv2d
Learn (contrib):
- BaseEstimator
- DNNClassifier
- DNNRegressor
- Estimator
- Evaluable
- evaluate
- extract_dask_data
- extract_dask_labels
- extract_pandas_data
- extract_pandas_labels
- extract_pandas_matrix
- infer
- LinearClassifier
- LinearRegressor
- LogisticRegressor
- ModeKeys
- NanLossDuringTrainingError
- read_batch_examples
- read_batch_features
- read_batch_record_features
- run_feeds
- run_n
- RunConfig
- train
- Trainable
Monitors (contrib):
- BaseMonitor
- CaptureVariable
- CheckpointSaver
- EveryN
- ExportMonitor
- get_default_monitors
- GraphDump
- LoggingTrainable
- NanLoss
- PrintTensor
- replace_monitors_with_hooks
- RunHookAdapterForMonitors
- StepCounter
- StopAtStep
- SummarySaver
- ValidationMonitor
Losses (contrib):
- absolute_difference
- add_loss
- compute_weighted_loss
- cosine_distance
- get_losses
- get_regularization_losses
- get_total_loss
- hinge_loss
- log_loss
- mean_pairwise_squared_error
- mean_squared_error
- sigmoid_cross_entropy
- softmax_cross_entropy
- sparse_softmax_cross_entropy
RNN (contrib):
- AttentionCellWrapper
- BidirectionalGridLSTMCell
- CoupledInputForgetGateLSTMCell
- FusedRNNCell
- FusedRNNCellAdaptor
- GridLSTMCell
- GRUBlockCell
- LayerNormBasicLSTMCell
- LSTMBlockCell
- LSTMBlockFusedCell
- LSTMBlockWrapper
- stack_bidirectional_dynamic_rnn
- stack_bidirectional_rnn
- TimeFreqLSTMCell
- TimeReversedFusedRNN
Metrics (contrib):
- accuracy
- aggregate_metric_map
- aggregate_metrics
- auc_using_histogram
- confusion_matrix
- set_difference
- set_intersection
- set_size
- set_union
- streaming_accuracy
- streaming_auc
- streaming_concat
- streaming_covariance
- streaming_mean
- streaming_mean_absolute_error
- streaming_mean_cosine_distance
- streaming_mean_iou
- streaming_mean_relative_error
- streaming_mean_squared_error
- streaming_pearson_correlation
- streaming_percentage_less
- streaming_precision
- streaming_recall
- streaming_recall_at_k
- streaming_root_mean_squared_error
- streaming_sensitivity_at_specificity
- streaming_sparse_average_precision_at_k
- streaming_sparse_precision_at_k
- streaming_sparse_precision_at_top_k
- streaming_sparse_recall_at_k
- streaming_specificity_at_sensitivity
Training (contrib):
- batch_sequences_with_states
- bucket
- bucket_by_sequence_length
- NextQueuedSequenceBatch
- rejection_sample
- resample_at_rate
- SequenceQueueingStateSaver
- stratified_sample
- weighted_resample
Utilities (contrib):
- constant_value
- make_ndarray
- make_tensor_proto
- ops_used_by_graph_def
- stripped_op_list_for_graph
Copying Graph Elements (contrib):
- copy_op_to_graph
- copy_variable_to_graph
- get_copied_op