Function reference • keras3

Models

Create Models

keras_model_sequential(): Keras Model composed of a linear stack of layers

keras_model(): Keras Model (Functional API)

keras_input(): Create a Keras tensor (Functional API input).

clone_model(): Clone a model instance.

Model(): Subclass the base Keras Model Class

Train Models

compile(<keras.src.models.model.Model>): Configure a model for training.

fit(<keras.src.models.model.Model>): Train a model for a fixed number of epochs (dataset iterations).

plot(<keras_training_history>): Plot training history

predict(<keras.src.models.model.Model>): Generates output predictions for the input samples.

evaluate(<keras.src.models.model.Model>): Evaluate a Keras Model

train_on_batch(): Runs a single gradient update on a single batch of data.

predict_on_batch(): Returns predictions for a single batch of samples.

test_on_batch(): Test the model on a single batch of samples.

freeze_weights() unfreeze_weights(): Freeze and unfreeze weights

Inspect and Modify Models

summary(<keras.src.models.model.Model>) format(<keras.src.models.model.Model>) print(<keras.src.models.model.Model>): Print a summary of a Keras Model

plot(<keras.src.models.model.Model>): Plot a Keras model

get_config() from_config(): Layer/Model configuration

get_weights() set_weights(): Layer/Model weights as R arrays

get_layer(): Retrieves a layer based on either its name (unique) or index.

count_params(): Count the total number of scalars composing the weights.

pop_layer(): Remove the last layer in a Sequential model

quantize_weights(): Quantize the weights of a model.

Save and Load Models

save_model(): Saves a model as a .keras file.

load_model(): Loads a model saved via save_model().

save_model_weights(): Saves all layer weights to a .weights.h5 file.

load_model_weights(): Load weights from a file saved via save_model_weights().

save_model_config() load_model_config(): Save and load model configuration as JSON

export_savedmodel(<keras.src.models.model.Model>): Create a TF SavedModel artifact for inference (e.g. via TF-Serving).

layer_tfsm(): Reload a Keras model/layer that was saved via export_savedmodel().

register_keras_serializable(): Registers a custom object with the Keras serialization framework.

Layers

Core Layers

layer_dense(): Just your regular densely-connected NN layer.

layer_einsum_dense(): A layer that uses einsum as the backing computation.

layer_embedding(): Turns positive integers (indexes) into dense vectors of fixed size.

layer_identity(): Identity layer.

layer_lambda(): Wraps arbitrary expressions as a Layer object.

layer_masking(): Masks a sequence by using a mask value to skip timesteps.

Reshaping Layers

layer_cropping_1d(): Cropping layer for 1D input (e.g. temporal sequence).

layer_cropping_2d(): Cropping layer for 2D input (e.g. picture).

layer_cropping_3d(): Cropping layer for 3D data (e.g. spatial or spatio-temporal).

layer_flatten(): Flattens the input. Does not affect the batch size.

layer_permute(): Permutes the dimensions of the input according to a given pattern.

layer_repeat_vector(): Repeats the input n times.

layer_reshape(): Layer that reshapes inputs into the given shape.

layer_upsampling_1d(): Upsampling layer for 1D inputs.

layer_upsampling_2d(): Upsampling layer for 2D inputs.

layer_upsampling_3d(): Upsampling layer for 3D inputs.

layer_zero_padding_1d(): Zero-padding layer for 1D input (e.g. temporal sequence).

layer_zero_padding_2d(): Zero-padding layer for 2D input (e.g. picture).

layer_zero_padding_3d(): Zero-padding layer for 3D data (spatial or spatio-temporal).

Convolutional Layers

layer_conv_1d(): 1D convolution layer (e.g. temporal convolution).

layer_conv_1d_transpose(): 1D transposed convolution layer.

layer_conv_2d(): 2D convolution layer.

layer_conv_2d_transpose(): 2D transposed convolution layer.

layer_conv_3d(): 3D convolution layer.

layer_conv_3d_transpose(): 3D transposed convolution layer.

layer_depthwise_conv_1d(): 1D depthwise convolution layer.

layer_depthwise_conv_2d(): 2D depthwise convolution layer.

layer_separable_conv_1d(): 1D separable convolution layer.

layer_separable_conv_2d(): 2D separable convolution layer.

Pooling Layers

layer_average_pooling_1d(): Average pooling for temporal data.

layer_average_pooling_2d(): Average pooling operation for 2D spatial data.

layer_average_pooling_3d(): Average pooling operation for 3D data (spatial or spatio-temporal).

layer_global_average_pooling_1d(): Global average pooling operation for temporal data.

layer_global_average_pooling_2d(): Global average pooling operation for 2D data.

layer_global_average_pooling_3d(): Global average pooling operation for 3D data.

layer_global_max_pooling_1d(): Global max pooling operation for temporal data.

layer_global_max_pooling_2d(): Global max pooling operation for 2D data.

layer_global_max_pooling_3d(): Global max pooling operation for 3D data.

layer_max_pooling_1d(): Max pooling operation for 1D temporal data.

layer_max_pooling_2d(): Max pooling operation for 2D spatial data.

layer_max_pooling_3d(): Max pooling operation for 3D data (spatial or spatio-temporal).

Activation Layers

layer_activation(): Applies an activation function to an output.

layer_activation_elu(): Applies an Exponential Linear Unit function to an output.

layer_activation_leaky_relu(): Leaky version of a Rectified Linear Unit activation layer.

layer_activation_parametric_relu(): Parametric Rectified Linear Unit activation layer.

layer_activation_relu(): Rectified Linear Unit activation function layer.

layer_activation_softmax(): Softmax activation layer.

Recurrent Layers

layer_bidirectional(): Bidirectional wrapper for RNNs.

layer_conv_lstm_1d(): 1D Convolutional LSTM.

layer_conv_lstm_2d(): 2D Convolutional LSTM.

layer_conv_lstm_3d(): 3D Convolutional LSTM.

layer_gru(): Gated Recurrent Unit - Cho et al. 2014.

layer_lstm(): Long Short-Term Memory layer - Hochreiter 1997.

layer_rnn(): Base class for recurrent layers

layer_simple_rnn(): Fully-connected RNN where the output is to be fed back as the new input.

layer_time_distributed(): This wrapper allows to apply a layer to every temporal slice of an input.

rnn_cell_gru(): Cell class for the GRU layer.

rnn_cell_lstm(): Cell class for the LSTM layer.

rnn_cell_simple(): Cell class for SimpleRNN.

rnn_cells_stack(): Wrapper allowing a stack of RNN cells to behave as a single cell.

reset_state(): Reset the state for a model, layer or metric.

Attention Layers

layer_additive_attention(): Additive attention layer, a.k.a. Bahdanau-style attention.

layer_attention(): Dot-product attention layer, a.k.a. Luong-style attention.

layer_group_query_attention(): Grouped Query Attention layer.

layer_multi_head_attention(): Multi Head Attention layer.

Normalization Layers

layer_batch_normalization(): Layer that normalizes its inputs.

layer_group_normalization(): Group normalization layer.

layer_layer_normalization(): Layer normalization layer (Ba et al., 2016).

layer_spectral_normalization(): Performs spectral normalization on the weights of a target layer.

layer_unit_normalization(): Unit normalization layer.

Regularization Layers

layer_activity_regularization(): Layer that applies an update to the cost function based input activity.

layer_alpha_dropout(): Applies Alpha Dropout to the input.

layer_dropout(): Applies dropout to the input.

layer_gaussian_dropout(): Apply multiplicative 1-centered Gaussian noise.

layer_gaussian_noise(): Apply additive zero-centered Gaussian noise.

layer_spatial_dropout_1d(): Spatial 1D version of Dropout.

layer_spatial_dropout_2d(): Spatial 2D version of Dropout.

layer_spatial_dropout_3d(): Spatial 3D version of Dropout.

Merging Layers

layer_add(): Performs elementwise addition operation.

layer_average(): Averages a list of inputs element-wise..

layer_concatenate(): Concatenates a list of inputs.

layer_dot(): Computes element-wise dot product of two tensors.

layer_maximum(): Computes element-wise maximum on a list of inputs.

layer_minimum(): Computes elementwise minimum on a list of inputs.

layer_multiply(): Performs elementwise multiplication.

layer_subtract(): Performs elementwise subtraction.

Preprocessing Layers

layer_category_encoding(): A preprocessing layer which encodes integer features.

layer_center_crop(): A preprocessing layer which crops images.

layer_discretization(): A preprocessing layer which buckets continuous features by ranges.

layer_feature_space() feature_cross() feature_custom() feature_float() feature_float_rescaled() feature_float_normalized() feature_float_discretized() feature_integer_categorical() feature_string_categorical() feature_string_hashed() feature_integer_hashed(): One-stop utility for preprocessing and encoding structured data.

layer_hashed_crossing(): A preprocessing layer which crosses features using the "hashing trick".

layer_hashing(): A preprocessing layer which hashes and bins categorical features.

layer_integer_lookup(): A preprocessing layer that maps integers to (possibly encoded) indices.

layer_mel_spectrogram(): A preprocessing layer to convert raw audio signals to Mel spectrograms.

layer_normalization(): A preprocessing layer that normalizes continuous features.

layer_random_brightness(): A preprocessing layer which randomly adjusts brightness during training.

layer_random_contrast(): A preprocessing layer which randomly adjusts contrast during training.

layer_random_crop(): A preprocessing layer which randomly crops images during training.

layer_random_flip(): A preprocessing layer which randomly flips images during training.

layer_random_rotation(): A preprocessing layer which randomly rotates images during training.

layer_random_translation(): A preprocessing layer which randomly translates images during training.

layer_random_zoom(): A preprocessing layer which randomly zooms images during training.

layer_rescaling(): A preprocessing layer which rescales input values to a new range.

layer_resizing(): A preprocessing layer which resizes images.

layer_string_lookup(): A preprocessing layer that maps strings to (possibly encoded) indices.

layer_text_vectorization() get_vocabulary() set_vocabulary(): A preprocessing layer which maps text features to integer sequences.

adapt(): Fits the state of the preprocessing layer to the data being passed

Compatability Layers

layer_tfsm(): Reload a Keras model/layer that was saved via export_savedmodel().

layer_jax_model_wrapper(): Keras Layer that wraps a JAX model.

layer_flax_module_wrapper(): Keras Layer that wraps a Flax module.

layer_torch_module_wrapper(): Torch module wrapper layer.

Custom Layers

layer_lambda(): Wraps arbitrary expressions as a Layer object.

Layer(): Define a custom Layer class.

Layer Methods

get_config() from_config(): Layer/Model configuration

get_weights() set_weights(): Layer/Model weights as R arrays

count_params(): Count the total number of scalars composing the weights.

reset_state(): Reset the state for a model, layer or metric.

Callbacks

callback_model_checkpoint(): Callback to save the Keras model or model weights at some frequency.

callback_backup_and_restore(): Callback to back up and restore the training state.

callback_early_stopping(): Stop training when a monitored metric has stopped improving.

callback_terminate_on_nan(): Callback that terminates training when a NaN loss is encountered.

callback_learning_rate_scheduler(): Learning rate scheduler.

callback_reduce_lr_on_plateau(): Reduce learning rate when a metric has stopped improving.

callback_csv_logger(): Callback that streams epoch results to a CSV file.

callback_tensorboard(): Enable visualizations for TensorBoard.

callback_remote_monitor(): Callback used to stream events to a server.

callback_lambda(): Callback for creating simple, custom callbacks on-the-fly.

callback_swap_ema_weights(): Swaps model weights and EMA weights before and after evaluation.

Callback(): Define a custom Callback class

Operations

Functions that are safe to call with both symbolic and eager tensor.

Core Operations

op_cast(): Cast a tensor to the desired dtype.

op_cond(): Conditionally applies true_fn or false_fn.

op_convert_to_numpy(): Convert a tensor to a NumPy array.

op_convert_to_tensor(): Convert an array to a tensor.

op_custom_gradient(): Decorator to define a function with a custom gradient.

op_fori_loop(): For loop implementation.

op_is_tensor(): Check whether the given object is a tensor.

op_scatter(): Returns a tensor of shape shape where indices are set to values.

op_scatter_update(): Update inputs via updates at scattered (sparse) indices.

op_shape(): Gets the shape of the tensor input.

op_slice(): Return a slice of an input tensor.

op_slice_update(): Update an input by slicing in a tensor of updated values.

op_stop_gradient(): Stops gradient computation.

op_unstack(): Unpacks the given dimension of a rank-R tensor into rank-(R-1) tensors.

op_vectorized_map(): Parallel map of function f on the first axis of tensor(s) elements.

op_while_loop(): While loop implementation.

Math Operations

op_erf(): Computes the error function of x, element-wise.

op_erfinv(): Computes the inverse error function of x, element-wise.

op_extract_sequences(): Expands the dimension of last axis into sequences of sequence_length.

op_fft(): Computes the Fast Fourier Transform along last axis of input.

op_fft2(): Computes the 2D Fast Fourier Transform along the last two axes of input.

op_in_top_k(): Checks if the targets are in the top-k predictions.

op_irfft(): Inverse real-valued Fast Fourier transform along the last axis.

op_istft(): Inverse Short-Time Fourier Transform along the last axis of the input.

op_logsumexp(): Computes the logarithm of sum of exponentials of elements in a tensor.

op_qr(): Computes the QR decomposition of a tensor.

op_rfft(): Real-valued Fast Fourier Transform along the last axis of the input.

op_rsqrt(): Computes reciprocal of square root of x element-wise.

op_segment_max(): Computes the max of segments in a tensor.

op_segment_sum(): Computes the sum of segments in a tensor.

op_solve(): Solves a linear system of equations given by a x = b.

op_stft(): Short-Time Fourier Transform along the last axis of the input.

op_top_k(): Finds the top-k values and their indices in a tensor.

General Tensor Operations

op_abs(): Compute the absolute value element-wise.

op_add(): Add arguments element-wise.

op_all(): Test whether all array elements along a given axis evaluate to TRUE.

op_any(): Test whether any array element along a given axis evaluates to TRUE.

op_append(): Append tensor x2 to the end of tensor x1.

op_arange(): Return evenly spaced values within a given interval.

op_arccos(): Trigonometric inverse cosine, element-wise.

op_arccosh(): Inverse hyperbolic cosine, element-wise.

op_arcsin(): Inverse sine, element-wise.

op_arcsinh(): Inverse hyperbolic sine, element-wise.

op_arctan(): Trigonometric inverse tangent, element-wise.

op_arctan2(): Element-wise arc tangent of x1/x2 choosing the quadrant correctly.

op_arctanh(): Inverse hyperbolic tangent, element-wise.

op_argmax(): Returns the indices of the maximum values along an axis.

op_argmin(): Returns the indices of the minimum values along an axis.

op_argsort(): Returns the indices that would sort a tensor.

op_array(): Create a tensor.

op_average(): Compute the weighted average along the specified axis.

op_bincount(): Count the number of occurrences of each value in a tensor of integers.

op_broadcast_to(): Broadcast a tensor to a new shape.

op_ceil(): Return the ceiling of the input, element-wise.

op_clip(): Clip (limit) the values in a tensor.

op_concatenate(): Join a sequence of tensors along an existing axis.

op_conj(): Returns the complex conjugate, element-wise.

op_copy(): Returns a copy of x.

op_correlate(): Compute the cross-correlation of two 1-dimensional tensors.

op_cos(): Cosine, element-wise.

op_cosh(): Hyperbolic cosine, element-wise.

op_count_nonzero(): Counts the number of non-zero values in x along the given axis.

op_cross(): Returns the cross product of two (arrays of) vectors.

op_ctc_decode(): Decodes the output of a CTC model.

op_cumprod(): Return the cumulative product of elements along a given axis.

op_cumsum(): Returns the cumulative sum of elements along a given axis.

op_diag(): Extract a diagonal or construct a diagonal array.

op_diagonal(): Return specified diagonals.

op_diff(): Calculate the n-th discrete difference along the given axis.

op_digitize(): Returns the indices of the bins to which each value in x belongs.

op_divide(): Divide arguments element-wise.

op_divide_no_nan(): Safe element-wise division which returns 0 where the denominator is 0.

op_dot(): Dot product of two tensors.

op_einsum(): Evaluates the Einstein summation convention on the operands.

op_empty(): Return a tensor of given shape and type filled with uninitialized data.

op_equal(): Returns (x1 == x2) element-wise.

op_exp(): Calculate the exponential of all elements in the input tensor.

op_expand_dims(): Expand the shape of a tensor.

op_expm1(): Calculate exp(x) - 1 for all elements in the tensor.

op_eye(): Return a 2-D tensor with ones on the diagonal and zeros elsewhere.

op_flip(): Reverse the order of elements in the tensor along the given axis.

op_floor(): Return the floor of the input, element-wise.

op_floor_divide(): Returns the largest integer smaller or equal to the division of inputs.

op_full(): Return a new tensor of given shape and type, filled with fill_value.

op_full_like(): Return a full tensor with the same shape and type as the given tensor.

op_get_item(): Return x[key].

op_greater(): Return the truth value of x1 > x2 element-wise.

op_greater_equal(): Return the truth value of x1 >= x2 element-wise.

op_hstack(): Stack tensors in sequence horizontally (column wise).

op_identity(): Return the identity tensor.

op_imag(): Return the imaginary part of the complex argument.

op_isclose(): Return whether two tensors are element-wise almost equal.

op_isfinite(): Return whether a tensor is finite, element-wise.

op_isinf(): Test element-wise for positive or negative infinity.

op_isnan(): Test element-wise for NaN and return result as a boolean tensor.

op_less(): Return the truth value of x1 < x2 element-wise.

op_less_equal(): Return the truth value of x1 <= x2 element-wise.

op_linspace(): Return evenly spaced numbers over a specified interval.

op_log(): Natural logarithm, element-wise.

op_log10(): Return the base 10 logarithm of the input tensor, element-wise.

op_log1p(): Returns the natural logarithm of one plus the x, element-wise.

op_log2(): Base-2 logarithm of x, element-wise.

op_logaddexp(): Logarithm of the sum of exponentiations of the inputs.

op_logical_and(): Computes the element-wise logical AND of the given input tensors.

op_logical_not(): Computes the element-wise NOT of the given input tensor.

op_logical_or(): Computes the element-wise logical OR of the given input tensors.

op_logical_xor(): Compute the truth value of x1 XOR x2, element-wise.

op_logspace(): Returns numbers spaced evenly on a log scale.

op_matmul(): Matrix product of two tensors.

op_max(): Return the maximum of a tensor or maximum along an axis.

op_maximum() op_pmax(): Element-wise maximum of x1 and x2.

op_mean(): Compute the arithmetic mean along the specified axes.

op_median(): Compute the median along the specified axis.

op_meshgrid(): Creates grids of coordinates from coordinate vectors.

op_min(): Return the minimum of a tensor or minimum along an axis.

op_minimum() op_pmin(): Element-wise minimum of x1 and x2.

op_mod(): Returns the element-wise remainder of division.

op_moveaxis(): Move axes of a tensor to new positions.

op_multiply(): Multiply arguments element-wise.

op_nan_to_num(): Replace NaN with zero and infinity with large finite numbers.

op_ndim(): Return the number of dimensions of a tensor.

op_negative(): Numerical negative, element-wise.

op_nonzero(): Return the indices of the elements that are non-zero.

op_not_equal(): Return (x1 != x2) element-wise.

op_ones(): Return a new tensor of given shape and type, filled with ones.

op_ones_like(): Return a tensor of ones with the same shape and type of x.

op_outer(): Compute the outer product of two vectors.

op_pad(): Pad a tensor.

op_power(): First tensor elements raised to powers from second tensor, element-wise.

op_prod(): Return the product of tensor elements over a given axis.

op_quantile(): Compute the q-th quantile(s) of the data along the specified axis.

op_ravel(): Return a contiguous flattened tensor.

op_real(): Return the real part of the complex argument.

op_reciprocal(): Return the reciprocal of the argument, element-wise.

op_repeat(): Repeat each element of a tensor after themselves.

op_reshape(): Gives a new shape to a tensor without changing its data.

op_roll(): Roll tensor elements along a given axis.

op_round(): Evenly round to the given number of decimals.

op_select(): Return elements from choicelist, based on conditions in condlist.

op_sign(): Returns a tensor with the signs of the elements of x.

op_sin(): Trigonometric sine, element-wise.

op_sinh(): Hyperbolic sine, element-wise.

op_size(): Return the number of elements in a tensor.

op_sort(): Sorts the elements of x along a given axis in ascending order.

op_split(): Split a tensor into chunks.

op_sqrt(): Return the non-negative square root of a tensor, element-wise.

op_square(): Return the element-wise square of the input.

op_squeeze(): Remove axes of length one from x.

op_stack(): Join a sequence of tensors along a new axis.

op_std(): Compute the standard deviation along the specified axis.

op_subtract(): Subtract arguments element-wise.

op_sum(): Sum of a tensor over the given axes.

op_swapaxes(): Interchange two axes of a tensor.

op_take(): Take elements from a tensor along an axis.

op_take_along_axis(): Select values from x at the 1-D indices along the given axis.

op_tan(): Compute tangent, element-wise.

op_tanh(): Hyperbolic tangent, element-wise.

op_tensordot(): Compute the tensor dot product along specified axes.

op_tile(): Repeat x the number of times given by repeats.

op_trace(): Return the sum along diagonals of the tensor.

op_transpose(): Returns a tensor with axes transposed.

op_tri(): Return a tensor with ones at and below a diagonal and zeros elsewhere.

op_tril(): Return lower triangle of a tensor.

op_triu(): Return upper triangle of a tensor.

op_var(): Compute the variance along the specified axes.

op_vdot(): Return the dot product of two vectors.

op_vectorize(): Turn a function into a vectorized function.

op_vstack(): Stack tensors in sequence vertically (row wise).

op_where(): Return elements chosen from x1 or x2 depending on condition.

op_zeros(): Return a new tensor of given shape and type, filled with zeros.

op_zeros_like(): Return a tensor of zeros with the same shape and type as x.

Neural Network Operations

op_average_pool(): Average pooling operation.

op_batch_normalization(): Normalizes x by mean and variance.

op_binary_crossentropy(): Computes binary cross-entropy loss between target and output tensor.

op_categorical_crossentropy(): Computes categorical cross-entropy loss between target and output tensor.

op_conv(): General N-D convolution.

op_conv_transpose(): General N-D convolution transpose.

op_ctc_loss(): CTC (Connectionist Temporal Classification) loss.

op_depthwise_conv(): General N-D depthwise convolution.

op_elu(): Exponential Linear Unit activation function.

op_gelu(): Gaussian Error Linear Unit (GELU) activation function.

op_hard_sigmoid(): Hard sigmoid activation function.

op_hard_silu() op_hard_swish(): Hard SiLU activation function, also known as Hard Swish.

op_leaky_relu(): Leaky version of a Rectified Linear Unit activation function.

op_log_sigmoid(): Logarithm of the sigmoid activation function.

op_log_softmax(): Log-softmax activation function.

op_max_pool(): Max pooling operation.

op_moments(): Calculates the mean and variance of x.

op_multi_hot(): Encodes integer labels as multi-hot vectors.

op_normalize(): Normalizes x over the specified axis.

op_one_hot(): Converts integer tensor x into a one-hot tensor.

op_relu(): Rectified linear unit activation function.

op_relu6(): Rectified linear unit activation function with upper bound of 6.

op_selu(): Scaled Exponential Linear Unit (SELU) activation function.

op_separable_conv(): General N-D separable convolution.

op_sigmoid(): Sigmoid activation function.

op_silu(): Sigmoid Linear Unit (SiLU) activation function, also known as Swish.

op_softmax(): Softmax activation function.

op_softplus(): Softplus activation function.

op_softsign(): Softsign activation function.

op_sparse_categorical_crossentropy(): Computes sparse categorical cross-entropy loss.

Linear Algebra Operations

op_cholesky(): Computes the Cholesky decomposition of a positive semi-definite matrix.

op_det(): Computes the determinant of a square tensor.

op_eig(): Computes the eigenvalues and eigenvectors of a square matrix.

op_eigh(): Computes the eigenvalues and eigenvectors of a complex Hermitian.

op_inv(): Computes the inverse of a square tensor.

op_lu_factor(): Computes the lower-upper decomposition of a square matrix.

op_norm(): Matrix or vector norm.

op_solve_triangular(): Solves a linear system of equations given by a %*% x = b.

op_svd(): Computes the singular value decomposition of a matrix.

Image Operations

op_image_affine_transform(): Applies the given transform(s) to the image(s).

op_image_crop(): Crop images to a specified height and width.

op_image_extract_patches(): Extracts patches from the image(s).

op_image_map_coordinates(): Map the input array to new coordinates by interpolation..

op_image_pad(): Pad images with zeros to the specified height and width.

op_image_resize(): Resize images to size using the specified interpolation method.

op_image_rgb_to_grayscale(): Convert RGB images to grayscale.

Losses

loss_binary_crossentropy(): Computes the cross-entropy loss between true labels and predicted labels.

loss_binary_focal_crossentropy(): Computes focal cross-entropy loss between true labels and predictions.

loss_categorical_crossentropy(): Computes the crossentropy loss between the labels and predictions.

loss_categorical_focal_crossentropy(): Computes the alpha balanced focal crossentropy loss.

loss_categorical_hinge(): Computes the categorical hinge loss between y_true & y_pred.

loss_cosine_similarity(): Computes the cosine similarity between y_true & y_pred.

loss_ctc(): CTC (Connectionist Temporal Classification) loss.

loss_dice(): Computes the Dice loss value between y_true and y_pred.

loss_hinge(): Computes the hinge loss between y_true & y_pred.

loss_huber(): Computes the Huber loss between y_true & y_pred.

loss_kl_divergence(): Computes Kullback-Leibler divergence loss between y_true & y_pred.

loss_log_cosh(): Computes the logarithm of the hyperbolic cosine of the prediction error.

loss_mean_absolute_error(): Computes the mean of absolute difference between labels and predictions.

loss_mean_absolute_percentage_error(): Computes the mean absolute percentage error between y_true and y_pred.

loss_mean_squared_error(): Computes the mean of squares of errors between labels and predictions.

loss_mean_squared_logarithmic_error(): Computes the mean squared logarithmic error between y_true and y_pred.

loss_poisson(): Computes the Poisson loss between y_true & y_pred.

loss_sparse_categorical_crossentropy(): Computes the crossentropy loss between the labels and predictions.

loss_squared_hinge(): Computes the squared hinge loss between y_true & y_pred.

loss_tversky(): Computes the Tversky loss value between y_true and y_pred.

Loss(): Subclass the base Loss class

Metrics

metric_auc(): Approximates the AUC (Area under the curve) of the ROC or PR curves.

metric_binary_accuracy(): Calculates how often predictions match binary labels.

metric_binary_crossentropy(): Computes the crossentropy metric between the labels and predictions.

metric_binary_focal_crossentropy(): Computes the binary focal crossentropy loss.

metric_binary_iou(): Computes the Intersection-Over-Union metric for class 0 and/or 1.

metric_categorical_accuracy(): Calculates how often predictions match one-hot labels.

metric_categorical_crossentropy(): Computes the crossentropy metric between the labels and predictions.

metric_categorical_focal_crossentropy(): Computes the categorical focal crossentropy loss.

metric_categorical_hinge(): Computes the categorical hinge metric between y_true and y_pred.

metric_cosine_similarity(): Computes the cosine similarity between the labels and predictions.

metric_f1_score(): Computes F-1 Score.

metric_false_negatives(): Calculates the number of false negatives.

metric_false_positives(): Calculates the number of false positives.

metric_fbeta_score(): Computes F-Beta score.

metric_hinge(): Computes the hinge metric between y_true and y_pred.

metric_huber(): Computes Huber loss value.

metric_iou(): Computes the Intersection-Over-Union metric for specific target classes.

metric_kl_divergence(): Computes Kullback-Leibler divergence metric between y_true and

metric_log_cosh(): Logarithm of the hyperbolic cosine of the prediction error.

metric_log_cosh_error(): Computes the logarithm of the hyperbolic cosine of the prediction error.

metric_mean(): Compute the (weighted) mean of the given values.

metric_mean_absolute_error(): Computes the mean absolute error between the labels and predictions.

metric_mean_absolute_percentage_error(): Computes mean absolute percentage error between y_true and y_pred.

metric_mean_iou(): Computes the mean Intersection-Over-Union metric.

metric_mean_squared_error(): Computes the mean squared error between y_true and y_pred.

metric_mean_squared_logarithmic_error(): Computes mean squared logarithmic error between y_true and y_pred.

metric_mean_wrapper(): Wrap a stateless metric function with the Mean metric.

metric_one_hot_iou(): Computes the Intersection-Over-Union metric for one-hot encoded labels.

metric_one_hot_mean_iou(): Computes mean Intersection-Over-Union metric for one-hot encoded labels.

metric_poisson(): Computes the Poisson metric between y_true and y_pred.

metric_precision(): Computes the precision of the predictions with respect to the labels.

metric_precision_at_recall(): Computes best precision where recall is >= specified value.

metric_r2_score(): Computes R2 score.

metric_recall(): Computes the recall of the predictions with respect to the labels.

metric_recall_at_precision(): Computes best recall where precision is >= specified value.

metric_root_mean_squared_error(): Computes root mean squared error metric between y_true and y_pred.

metric_sensitivity_at_specificity(): Computes best sensitivity where specificity is >= specified value.

metric_sparse_categorical_accuracy(): Calculates how often predictions match integer labels.

metric_sparse_categorical_crossentropy(): Computes the crossentropy metric between the labels and predictions.

metric_sparse_top_k_categorical_accuracy(): Computes how often integer targets are in the top K predictions.

metric_specificity_at_sensitivity(): Computes best specificity where sensitivity is >= specified value.

metric_squared_hinge(): Computes the hinge metric between y_true and y_pred.

metric_sum(): Compute the (weighted) sum of the given values.

metric_top_k_categorical_accuracy(): Computes how often targets are in the top K predictions.

metric_true_negatives(): Calculates the number of true negatives.

metric_true_positives(): Calculates the number of true positives.

custom_metric(): Custom metric function

reset_state(): Reset the state for a model, layer or metric.

Metric(): Subclass the base Metric class

Data Loading

Keras data loading utilities help you quickly go from raw data to a TF Dataset object that can be used to efficiently train a model. These loading utilites can be combined with preprocessing layers to futher transform your input dataset before training.

image_dataset_from_directory(): Generates a tf.data.Dataset from image files in a directory.

text_dataset_from_directory(): Generates a tf.data.Dataset from text files in a directory.

audio_dataset_from_directory(): Generates a tf.data.Dataset from audio files in a directory.

timeseries_dataset_from_array(): Creates a dataset of sliding windows over a timeseries provided as array.

Preprocessing

layer_feature_space() feature_cross() feature_custom() feature_float() feature_float_rescaled() feature_float_normalized() feature_float_discretized() feature_integer_categorical() feature_string_categorical() feature_string_hashed() feature_integer_hashed(): One-stop utility for preprocessing and encoding structured data.

adapt(): Fits the state of the preprocessing layer to the data being passed

Numerical Features Preprocessing Layers

layer_normalization(): A preprocessing layer that normalizes continuous features.

layer_discretization(): A preprocessing layer which buckets continuous features by ranges.

Categorical Features Preprocessing Layers

layer_category_encoding(): A preprocessing layer which encodes integer features.

layer_hashing(): A preprocessing layer which hashes and bins categorical features.

layer_hashed_crossing(): A preprocessing layer which crosses features using the "hashing trick".

layer_string_lookup(): A preprocessing layer that maps strings to (possibly encoded) indices.

layer_integer_lookup(): A preprocessing layer that maps integers to (possibly encoded) indices.

Text Preprocessing Layers

layer_text_vectorization() get_vocabulary() set_vocabulary(): A preprocessing layer which maps text features to integer sequences.

Sequence Preprocessing

timeseries_dataset_from_array(): Creates a dataset of sliding windows over a timeseries provided as array.

pad_sequences(): Pads sequences to the same length.

Image Preprocessing Layers

layer_resizing(): A preprocessing layer which resizes images.

layer_rescaling(): A preprocessing layer which rescales input values to a new range.

layer_center_crop(): A preprocessing layer which crops images.

Image Preprocessing

image_array_save(): Saves an image stored as an array to a path or file object.

image_dataset_from_directory(): Generates a tf.data.Dataset from image files in a directory.

image_from_array(): Converts a 3D array to a PIL Image instance.

image_load(): Loads an image into PIL format.

image_smart_resize(): Resize images to a target size without aspect ratio distortion.

image_to_array(): Converts a PIL Image instance to a matrix.

op_image_affine_transform(): Applies the given transform(s) to the image(s).

op_image_crop(): Crop images to a specified height and width.

op_image_extract_patches(): Extracts patches from the image(s).

op_image_map_coordinates(): Map the input array to new coordinates by interpolation..

op_image_pad(): Pad images with zeros to the specified height and width.

op_image_resize(): Resize images to size using the specified interpolation method.

op_image_rgb_to_grayscale(): Convert RGB images to grayscale.

Image augmentation Layers

layer_random_crop(): A preprocessing layer which randomly crops images during training.

layer_random_flip(): A preprocessing layer which randomly flips images during training.

layer_random_translation(): A preprocessing layer which randomly translates images during training.

layer_random_rotation(): A preprocessing layer which randomly rotates images during training.

layer_random_zoom(): A preprocessing layer which randomly zooms images during training.

layer_random_contrast(): A preprocessing layer which randomly adjusts contrast during training.

layer_random_brightness(): A preprocessing layer which randomly adjusts brightness during training.

Application Preprocessing

application_preprocess_inputs() application_decode_predictions(): Preprocessing and postprocessing utilities

Optimizers

optimizer_adadelta(): Optimizer that implements the Adadelta algorithm.

optimizer_adafactor(): Optimizer that implements the Adafactor algorithm.

optimizer_adagrad(): Optimizer that implements the Adagrad algorithm.

optimizer_adam(): Optimizer that implements the Adam algorithm.

optimizer_adam_w(): Optimizer that implements the AdamW algorithm.

optimizer_adamax(): Optimizer that implements the Adamax algorithm.

optimizer_ftrl(): Optimizer that implements the FTRL algorithm.

optimizer_lion(): Optimizer that implements the Lion algorithm.

optimizer_loss_scale(): An optimizer that dynamically scales the loss to prevent underflow.

optimizer_nadam(): Optimizer that implements the Nadam algorithm.

optimizer_rmsprop(): Optimizer that implements the RMSprop algorithm.

optimizer_sgd(): Gradient descent (with momentum) optimizer.

Learning Rate Schedules

learning_rate_schedule_cosine_decay(): A LearningRateSchedule that uses a cosine decay with optional warmup.

learning_rate_schedule_cosine_decay_restarts(): A LearningRateSchedule that uses a cosine decay schedule with restarts.

learning_rate_schedule_exponential_decay(): A LearningRateSchedule that uses an exponential decay schedule.

learning_rate_schedule_inverse_time_decay(): A LearningRateSchedule that uses an inverse time decay schedule.

learning_rate_schedule_piecewise_constant_decay(): A LearningRateSchedule that uses a piecewise constant decay schedule.

learning_rate_schedule_polynomial_decay(): A LearningRateSchedule that uses a polynomial decay schedule.

LearningRateSchedule(): Define a custom LearningRateSchedule class

Initializers

initializer_constant(): Initializer that generates tensors with constant values.

initializer_glorot_normal(): The Glorot normal initializer, also called Xavier normal initializer.

initializer_glorot_uniform(): The Glorot uniform initializer, also called Xavier uniform initializer.

initializer_he_normal(): He normal initializer.

initializer_he_uniform(): He uniform variance scaling initializer.

initializer_identity(): Initializer that generates the identity matrix.

initializer_lecun_normal(): Lecun normal initializer.

initializer_lecun_uniform(): Lecun uniform initializer.

initializer_ones(): Initializer that generates tensors initialized to 1.

initializer_orthogonal(): Initializer that generates an orthogonal matrix.

initializer_random_normal(): Random normal initializer.

initializer_random_uniform(): Random uniform initializer.

initializer_truncated_normal(): Initializer that generates a truncated normal distribution.

initializer_variance_scaling(): Initializer that adapts its scale to the shape of its input tensors.

initializer_zeros(): Initializer that generates tensors initialized to 0.

Constraints

Constraint(): Define a custom Constraint class

constraint_maxnorm(): MaxNorm weight constraint.

constraint_minmaxnorm(): MinMaxNorm weight constraint.

constraint_nonneg(): Constrains the weights to be non-negative.

constraint_unitnorm(): Constrains the weights incident to each hidden unit to have unit norm.

Regularizers

regularizer_l1(): A regularizer that applies a L1 regularization penalty.

regularizer_l1_l2(): A regularizer that applies both L1 and L2 regularization penalties.

regularizer_l2(): A regularizer that applies a L2 regularization penalty.

regularizer_orthogonal(): Regularizer that encourages input vectors to be orthogonal to each other.

Activations

activation_elu(): Exponential Linear Unit.

activation_exponential(): Exponential activation function.

activation_gelu(): Gaussian error linear unit (GELU) activation function.

activation_hard_sigmoid(): Hard sigmoid activation function.

activation_hard_silu() activation_hard_swish(): Hard SiLU activation function, also known as Hard Swish.

activation_leaky_relu(): Leaky relu activation function.

activation_linear(): Linear activation function (pass-through).

activation_log_softmax(): Log-Softmax activation function.

activation_mish(): Mish activation function.

activation_relu(): Applies the rectified linear unit activation function.

activation_relu6(): Relu6 activation function.

activation_selu(): Scaled Exponential Linear Unit (SELU).

activation_sigmoid(): Sigmoid activation function.

activation_silu(): Swish (or Silu) activation function.

activation_softmax(): Softmax converts a vector of values to a probability distribution.

activation_softplus(): Softplus activation function.

activation_softsign(): Softsign activation function.

activation_tanh(): Hyperbolic tangent activation function.

Random Tensor Generators

random_uniform(): Draw samples from a uniform distribution.

random_normal(): Draw random samples from a normal (Gaussian) distribution.

random_truncated_normal(): Draw samples from a truncated normal distribution.

random_gamma(): Draw random samples from the Gamma distribution.

random_categorical(): Draws samples from a categorical distribution.

random_integer(): Draw random integers from a uniform distribution.

random_dropout(): Randomly set some values in a tensor to 0.

random_shuffle(): Shuffle the elements of a tensor uniformly at random along an axis.

random_beta(): Draw samples from a Beta distribution.

random_binomial(): Draw samples from a Binomial distribution.

random_seed_generator(): Generates variable seeds upon each call to a RNG-using function.

Builtin small datasets

dataset_boston_housing(): Boston housing price regression dataset

dataset_cifar10(): CIFAR10 small image classification

dataset_cifar100(): CIFAR100 small image classification

dataset_fashion_mnist(): Fashion-MNIST database of fashion articles

dataset_imdb() dataset_imdb_word_index(): IMDB Movie reviews sentiment classification

dataset_mnist(): MNIST database of handwritten digits

dataset_reuters() dataset_reuters_word_index(): Reuters newswire topics classification

Configuration

config_backend(): Publicly accessible method for determining the current backend.

config_disable_interactive_logging(): Turn off interactive logging.

config_disable_traceback_filtering(): Turn off traceback filtering.

config_dtype_policy(): Returns the current default dtype policy object.

config_enable_interactive_logging(): Turn on interactive logging.

config_enable_traceback_filtering(): Turn on traceback filtering.

config_enable_unsafe_deserialization(): Disables safe mode globally, allowing deserialization of lambdas.

config_epsilon(): Return the value of the fuzz factor used in numeric expressions.

config_floatx(): Return the default float type, as a string.

config_image_data_format(): Return the default image data format convention.

config_is_interactive_logging_enabled(): Check if interactive logging is enabled.

config_is_traceback_filtering_enabled(): Check if traceback filtering is enabled.

config_set_backend(): Reload the backend (and the Keras package).

config_set_dtype_policy(): Sets the default dtype policy globally.

config_set_epsilon(): Set the value of the fuzz factor used in numeric expressions.

config_set_floatx(): Set the default float dtype.

config_set_image_data_format(): Set the value of the image data format convention.

Utils

install_keras(): Install Keras

use_backend(): Configure a Keras backend

shape() format(<keras_shape>) print(<keras_shape>) `[`(<keras_shape>) as.integer(<keras_shape>) as.list(<keras_shape>): Tensor shape utility

set_random_seed(): Sets all random seeds (Python, NumPy, and backend framework, e.g. TF).

clear_session(): Resets all state generated by Keras.

get_source_inputs(): Returns the list of input tensors necessary to compute tensor.

keras: Main Keras module

Numerical Utils

normalize(): Normalizes an array.

to_categorical(): Converts a class vector (integers) to binary class matrix.

Data Utils

zip_lists(): Zip lists

get_file(): Downloads a file from a URL if it not already in the cache.

split_dataset(): Splits a dataset into a left half and a right half (e.g. train / test).

pack_x_y_sample_weight(): Packs user-provided data into a list.

unpack_x_y_sample_weight(): Unpacks user-provided data list.

Serialization Utils

register_keras_serializable(): Registers a custom object with the Keras serialization framework.

get_custom_objects() set_custom_objects(): Get/set the currently registered custom objects.

get_registered_name(): Returns the name registered to an object within the Keras framework.

get_registered_object(): Returns the class associated with name if it is registered with Keras.

serialize_keras_object(): Retrieve the full config by serializing the Keras object.

deserialize_keras_object(): Retrieve the object by deserializing the config dict.

with_custom_object_scope(): Provide a scope with mappings of names to custom objects

config_enable_unsafe_deserialization(): Disables safe mode globally, allowing deserialization of lambdas.

Base Keras Classes

Define custom object by subclassing base Keras classes.

Layer(): Define a custom Layer class.

Loss(): Subclass the base Loss class

Metric(): Subclass the base Metric class

Callback(): Define a custom Callback class

Constraint(): Define a custom Constraint class

Model(): Subclass the base Keras Model Class

LearningRateSchedule(): Define a custom LearningRateSchedule class

active_property(): Create an active property class method

Applications

Application utilities

application_preprocess_inputs() application_decode_predictions(): Preprocessing and postprocessing utilities

ConvNeXt Applications

application_convnext_base(): Instantiates the ConvNeXtBase architecture.

application_convnext_large(): Instantiates the ConvNeXtLarge architecture.

application_convnext_small(): Instantiates the ConvNeXtSmall architecture.

application_convnext_tiny(): Instantiates the ConvNeXtTiny architecture.

application_convnext_xlarge(): Instantiates the ConvNeXtXLarge architecture.

Densenet Applications

application_densenet121(): Instantiates the Densenet121 architecture.

application_densenet169(): Instantiates the Densenet169 architecture.

application_densenet201(): Instantiates the Densenet201 architecture.

EfficientNet Applications

application_efficientnet_b0(): Instantiates the EfficientNetB0 architecture.

application_efficientnet_b1(): Instantiates the EfficientNetB1 architecture.

application_efficientnet_b2(): Instantiates the EfficientNetB2 architecture.

application_efficientnet_b3(): Instantiates the EfficientNetB3 architecture.

application_efficientnet_b4(): Instantiates the EfficientNetB4 architecture.

application_efficientnet_b5(): Instantiates the EfficientNetB5 architecture.

application_efficientnet_b6(): Instantiates the EfficientNetB6 architecture.

application_efficientnet_b7(): Instantiates the EfficientNetB7 architecture.

application_efficientnet_v2b0(): Instantiates the EfficientNetV2B0 architecture.

application_efficientnet_v2b1(): Instantiates the EfficientNetV2B1 architecture.

application_efficientnet_v2b2(): Instantiates the EfficientNetV2B2 architecture.

application_efficientnet_v2b3(): Instantiates the EfficientNetV2B3 architecture.

application_efficientnet_v2l(): Instantiates the EfficientNetV2L architecture.

application_efficientnet_v2m(): Instantiates the EfficientNetV2M architecture.

application_efficientnet_v2s(): Instantiates the EfficientNetV2S architecture.

Inception Applications

application_inception_resnet_v2(): Instantiates the Inception-ResNet v2 architecture.

application_inception_v3(): Instantiates the Inception v3 architecture.

MobileNet Applications

application_mobilenet(): Instantiates the MobileNet architecture.

application_mobilenet_v2(): Instantiates the MobileNetV2 architecture.

application_mobilenet_v3_large(): Instantiates the MobileNetV3Large architecture.

application_mobilenet_v3_small(): Instantiates the MobileNetV3Small architecture.

NASNet Applications

application_nasnetlarge(): Instantiates a NASNet model in ImageNet mode.

application_nasnetmobile(): Instantiates a Mobile NASNet model in ImageNet mode.

ResNet Applications

application_resnet101(): Instantiates the ResNet101 architecture.

application_resnet101_v2(): Instantiates the ResNet101V2 architecture.

application_resnet152(): Instantiates the ResNet152 architecture.

application_resnet152_v2(): Instantiates the ResNet152V2 architecture.

application_resnet50(): Instantiates the ResNet50 architecture.

application_resnet50_v2(): Instantiates the ResNet50V2 architecture.

VGG Applications

application_vgg16(): Instantiates the VGG16 model.

application_vgg19(): Instantiates the VGG19 model.

Xception Applications

application_xception(): Instantiates the Xception architecture.