zea.ops.keras_ops

Auto-generated zea.Operation for all unary keras.ops and keras.ops.image functions.

They can be used in zea pipelines like any other zea.Operation, for example:

>>> from zea.ops.keras_ops import Squeeze

>>> op = Squeeze(axis=1)

This file is generated automatically. Do not edit manually. Generated with Keras 3.14.0

Classes

Abs(*args, **kwargs)

Operation wrapping keras.ops.abs.

Absolute(*args, **kwargs)

Operation wrapping keras.ops.absolute.

AffineTransform(*args, **kwargs)

Operation wrapping keras.ops.image.affine_transform.

All(*args, **kwargs)

Operation wrapping keras.ops.all.

Amax(*args, **kwargs)

Operation wrapping keras.ops.amax.

Amin(*args, **kwargs)

Operation wrapping keras.ops.amin.

Angle(*args, **kwargs)

Operation wrapping keras.ops.angle.

Any(*args, **kwargs)

Operation wrapping keras.ops.any.

Arccos(*args, **kwargs)

Operation wrapping keras.ops.arccos.

Arccosh(*args, **kwargs)

Operation wrapping keras.ops.arccosh.

Arcsin(*args, **kwargs)

Operation wrapping keras.ops.arcsin.

Arcsinh(*args, **kwargs)

Operation wrapping keras.ops.arcsinh.

Arctan(*args, **kwargs)

Operation wrapping keras.ops.arctan.

Arctanh(*args, **kwargs)

Operation wrapping keras.ops.arctanh.

Argmax(*args, **kwargs)

Operation wrapping keras.ops.argmax.

Argmin(*args, **kwargs)

Operation wrapping keras.ops.argmin.

Argpartition(*args, **kwargs)

Operation wrapping keras.ops.argpartition.

Argsort(*args, **kwargs)

Operation wrapping keras.ops.argsort.

Array(*args, **kwargs)

Operation wrapping keras.ops.array.

ArraySplit(*args, **kwargs)

Operation wrapping keras.ops.array_split.

Average(*args, **kwargs)

Operation wrapping keras.ops.average.

Bartlett(*args, **kwargs)

Operation wrapping keras.ops.bartlett.

BatchNormalization(*args, **kwargs)

Operation wrapping keras.ops.batch_normalization.

Bincount(*args, **kwargs)

Operation wrapping keras.ops.bincount.

BitwiseAnd(*args, **kwargs)

Operation wrapping keras.ops.bitwise_and.

BitwiseInvert(*args, **kwargs)

Operation wrapping keras.ops.bitwise_invert.

BitwiseLeftShift(*args, **kwargs)

Operation wrapping keras.ops.bitwise_left_shift.

BitwiseNot(*args, **kwargs)

Operation wrapping keras.ops.bitwise_not.

BitwiseOr(*args, **kwargs)

Operation wrapping keras.ops.bitwise_or.

BitwiseRightShift(*args, **kwargs)

Operation wrapping keras.ops.bitwise_right_shift.

BitwiseXor(*args, **kwargs)

Operation wrapping keras.ops.bitwise_xor.

Blackman(*args, **kwargs)

Operation wrapping keras.ops.blackman.

BroadcastTo(*args, **kwargs)

Operation wrapping keras.ops.broadcast_to.

Cast(*args, **kwargs)

Operation wrapping keras.ops.cast.

Cbrt(*args, **kwargs)

Operation wrapping keras.ops.cbrt.

Ceil(*args, **kwargs)

Operation wrapping keras.ops.ceil.

Celu(*args, **kwargs)

Operation wrapping keras.ops.celu.

Cholesky(*args, **kwargs)

Operation wrapping keras.ops.cholesky.

CholeskyInverse(*args, **kwargs)

Operation wrapping keras.ops.cholesky_inverse.

Clip(*args, **kwargs)

Operation wrapping keras.ops.clip.

Conj(*args, **kwargs)

Operation wrapping keras.ops.conj.

Conjugate(*args, **kwargs)

Operation wrapping keras.ops.conjugate.

ConvertToNumpy(*args, **kwargs)

Operation wrapping keras.ops.convert_to_numpy.

ConvertToTensor(*args, **kwargs)

Operation wrapping keras.ops.convert_to_tensor.

Copy(*args, **kwargs)

Operation wrapping keras.ops.copy.

Corrcoef(*args, **kwargs)

Operation wrapping keras.ops.corrcoef.

Cos(*args, **kwargs)

Operation wrapping keras.ops.cos.

Cosh(*args, **kwargs)

Operation wrapping keras.ops.cosh.

CountNonzero(*args, **kwargs)

Operation wrapping keras.ops.count_nonzero.

CropImages(*args, **kwargs)

Operation wrapping keras.ops.image.crop_images.

Cumprod(*args, **kwargs)

Operation wrapping keras.ops.cumprod.

Cumsum(*args, **kwargs)

Operation wrapping keras.ops.cumsum.

Deg2rad(*args, **kwargs)

Operation wrapping keras.ops.deg2rad.

DepthToSpace(*args, **kwargs)

Operation wrapping keras.ops.depth_to_space.

Det(*args, **kwargs)

Operation wrapping keras.ops.det.

Diag(*args, **kwargs)

Operation wrapping keras.ops.diag.

Diagflat(*args, **kwargs)

Operation wrapping keras.ops.diagflat.

Diagonal(*args, **kwargs)

Operation wrapping keras.ops.diagonal.

Digitize(*args, **kwargs)

Operation wrapping keras.ops.digitize.

Dtype(*args, **kwargs)

Operation wrapping keras.ops.dtype.

Eig(*args, **kwargs)

Operation wrapping keras.ops.eig.

Eigh(*args, **kwargs)

Operation wrapping keras.ops.eigh.

ElasticTransform(*args, **kwargs)

Operation wrapping keras.ops.image.elastic_transform.

Elu(*args, **kwargs)

Operation wrapping keras.ops.elu.

EmptyLike(*args, **kwargs)

Operation wrapping keras.ops.empty_like.

Erf(*args, **kwargs)

Operation wrapping keras.ops.erf.

Erfinv(*args, **kwargs)

Operation wrapping keras.ops.erfinv.

Exp(*args, **kwargs)

Operation wrapping keras.ops.exp.

Exp2(*args, **kwargs)

Operation wrapping keras.ops.exp2.

ExpandDims(*args, **kwargs)

Operation wrapping keras.ops.expand_dims.

Expm1(*args, **kwargs)

Operation wrapping keras.ops.expm1.

ExtractPatches(*args, **kwargs)

Operation wrapping keras.ops.image.extract_patches.

ExtractSequences(*args, **kwargs)

Operation wrapping keras.ops.extract_sequences.

Fft(*args, **kwargs)

Operation wrapping keras.ops.fft.

Fft2(*args, **kwargs)

Operation wrapping keras.ops.fft2.

Flip(*args, **kwargs)

Operation wrapping keras.ops.flip.

Fliplr(*args, **kwargs)

Operation wrapping keras.ops.fliplr.

Flipud(*args, **kwargs)

Operation wrapping keras.ops.flipud.

Floor(*args, **kwargs)

Operation wrapping keras.ops.floor.

Fold(*args, **kwargs)

Operation wrapping keras.ops.fold.

FullLike(*args, **kwargs)

Operation wrapping keras.ops.full_like.

GaussianBlur(*args, **kwargs)

Operation wrapping keras.ops.image.gaussian_blur.

Gelu(*args, **kwargs)

Operation wrapping keras.ops.gelu.

GetItem(*args, **kwargs)

Operation wrapping keras.ops.get_item.

Glu(*args, **kwargs)

Operation wrapping keras.ops.glu.

Hamming(*args, **kwargs)

Operation wrapping keras.ops.hamming.

Hanning(*args, **kwargs)

Operation wrapping keras.ops.hanning.

HardShrink(*args, **kwargs)

Operation wrapping keras.ops.hard_shrink.

HardSigmoid(*args, **kwargs)

Operation wrapping keras.ops.hard_sigmoid.

HardSilu(*args, **kwargs)

Operation wrapping keras.ops.hard_silu.

HardSwish(*args, **kwargs)

Operation wrapping keras.ops.hard_swish.

HardTanh(*args, **kwargs)

Operation wrapping keras.ops.hard_tanh.

Histogram(*args, **kwargs)

Operation wrapping keras.ops.histogram.

Hsplit(*args, **kwargs)

Operation wrapping keras.ops.hsplit.

HsvToRgb(*args, **kwargs)

Operation wrapping keras.ops.image.hsv_to_rgb.

I0(*args, **kwargs)

Operation wrapping keras.ops.i0.

Ifft2(*args, **kwargs)

Operation wrapping keras.ops.ifft2.

Imag(*args, **kwargs)

Operation wrapping keras.ops.imag.

Inv(*args, **kwargs)

Operation wrapping keras.ops.inv.

Irfft(*args, **kwargs)

Operation wrapping keras.ops.irfft.

IsTensor(*args, **kwargs)

Operation wrapping keras.ops.is_tensor.

Isfinite(*args, **kwargs)

Operation wrapping keras.ops.isfinite.

Isinf(*args, **kwargs)

Operation wrapping keras.ops.isinf.

Isnan(*args, **kwargs)

Operation wrapping keras.ops.isnan.

Isneginf(*args, **kwargs)

Operation wrapping keras.ops.isneginf.

Isposinf(*args, **kwargs)

Operation wrapping keras.ops.isposinf.

Isreal(*args, **kwargs)

Operation wrapping keras.ops.isreal.

Istft(*args, **kwargs)

Operation wrapping keras.ops.istft.

Kaiser(*args, **kwargs)

Operation wrapping keras.ops.kaiser.

LayerNormalization(*args, **kwargs)

Operation wrapping keras.ops.layer_normalization.

LeakyRelu(*args, **kwargs)

Operation wrapping keras.ops.leaky_relu.

LeftShift(*args, **kwargs)

Operation wrapping keras.ops.left_shift.

Log(*args, **kwargs)

Operation wrapping keras.ops.log.

Log10(*args, **kwargs)

Operation wrapping keras.ops.log10.

Log1p(*args, **kwargs)

Operation wrapping keras.ops.log1p.

Log2(*args, **kwargs)

Operation wrapping keras.ops.log2.

LogSigmoid(*args, **kwargs)

Operation wrapping keras.ops.log_sigmoid.

LogSoftmax(*args, **kwargs)

Operation wrapping keras.ops.log_softmax.

Logdet(*args, **kwargs)

Operation wrapping keras.ops.logdet.

LogicalNot(*args, **kwargs)

Operation wrapping keras.ops.logical_not.

Logsumexp(*args, **kwargs)

Operation wrapping keras.ops.logsumexp.

LuFactor(*args, **kwargs)

Operation wrapping keras.ops.lu_factor.

Max(*args, **kwargs)

Operation wrapping keras.ops.max.

Mean(*args, **kwargs)

Operation wrapping keras.ops.mean.

Median(*args, **kwargs)

Operation wrapping keras.ops.median.

Meshgrid(*args, **kwargs)

Operation wrapping keras.ops.meshgrid.

Min(*args, **kwargs)

Operation wrapping keras.ops.min.

Moments(*args, **kwargs)

Operation wrapping keras.ops.moments.

Moveaxis(*args, **kwargs)

Operation wrapping keras.ops.moveaxis.

NanToNum(*args, **kwargs)

Operation wrapping keras.ops.nan_to_num.

Nanargmax(*args, **kwargs)

Operation wrapping keras.ops.nanargmax.

Nanargmin(*args, **kwargs)

Operation wrapping keras.ops.nanargmin.

Nancumprod(*args, **kwargs)

Operation wrapping keras.ops.nancumprod.

Nancumsum(*args, **kwargs)

Operation wrapping keras.ops.nancumsum.

Nanmax(*args, **kwargs)

Operation wrapping keras.ops.nanmax.

Nanmean(*args, **kwargs)

Operation wrapping keras.ops.nanmean.

Nanmedian(*args, **kwargs)

Operation wrapping keras.ops.nanmedian.

Nanmin(*args, **kwargs)

Operation wrapping keras.ops.nanmin.

Nanprod(*args, **kwargs)

Operation wrapping keras.ops.nanprod.

Nanquantile(*args, **kwargs)

Operation wrapping keras.ops.nanquantile.

Nanstd(*args, **kwargs)

Operation wrapping keras.ops.nanstd.

Nansum(*args, **kwargs)

Operation wrapping keras.ops.nansum.

Nanvar(*args, **kwargs)

Operation wrapping keras.ops.nanvar.

Ndim(*args, **kwargs)

Operation wrapping keras.ops.ndim.

Negative(*args, **kwargs)

Operation wrapping keras.ops.negative.

Nonzero(*args, **kwargs)

Operation wrapping keras.ops.nonzero.

Norm(*args, **kwargs)

Operation wrapping keras.ops.norm.

Normalize(*args, **kwargs)

Operation wrapping keras.ops.normalize.

OneHot(*args, **kwargs)

Operation wrapping keras.ops.one_hot.

OnesLike(*args, **kwargs)

Operation wrapping keras.ops.ones_like.

Pad(*args, **kwargs)

Operation wrapping keras.ops.pad.

PadImages(*args, **kwargs)

Operation wrapping keras.ops.image.pad_images.

PerspectiveTransform(*args, **kwargs)

Operation wrapping keras.ops.image.perspective_transform.

Prod(*args, **kwargs)

Operation wrapping keras.ops.prod.

Ptp(*args, **kwargs)

Operation wrapping keras.ops.ptp.

Qr(*args, **kwargs)

Operation wrapping keras.ops.qr.

Quantile(*args, **kwargs)

Operation wrapping keras.ops.quantile.

Rad2deg(*args, **kwargs)

Operation wrapping keras.ops.rad2deg.

Ravel(*args, **kwargs)

Operation wrapping keras.ops.ravel.

Real(*args, **kwargs)

Operation wrapping keras.ops.real.

Reciprocal(*args, **kwargs)

Operation wrapping keras.ops.reciprocal.

Relu(*args, **kwargs)

Operation wrapping keras.ops.relu.

Relu6(*args, **kwargs)

Operation wrapping keras.ops.relu6.

Repeat(*args, **kwargs)

Operation wrapping keras.ops.repeat.

Reshape(*args, **kwargs)

Operation wrapping keras.ops.reshape.

Resize(*args, **kwargs)

Operation wrapping keras.ops.image.resize.

Rfft(*args, **kwargs)

Operation wrapping keras.ops.rfft.

RgbToGrayscale(*args, **kwargs)

Operation wrapping keras.ops.image.rgb_to_grayscale.

RgbToHsv(*args, **kwargs)

Operation wrapping keras.ops.image.rgb_to_hsv.

RightShift(*args, **kwargs)

Operation wrapping keras.ops.right_shift.

RmsNormalization(*args, **kwargs)

Operation wrapping keras.ops.rms_normalization.

Roll(*args, **kwargs)

Operation wrapping keras.ops.roll.

Round(*args, **kwargs)

Operation wrapping keras.ops.round.

Rsqrt(*args, **kwargs)

Operation wrapping keras.ops.rsqrt.

SaturateCast(*args, **kwargs)

Operation wrapping keras.ops.saturate_cast.

ScaleAndTranslate(*args, **kwargs)

Operation wrapping keras.ops.image.scale_and_translate.

Selu(*args, **kwargs)

Operation wrapping keras.ops.selu.

Shape(*args, **kwargs)

Operation wrapping keras.ops.shape.

Sigmoid(*args, **kwargs)

Operation wrapping keras.ops.sigmoid.

Sign(*args, **kwargs)

Operation wrapping keras.ops.sign.

Signbit(*args, **kwargs)

Operation wrapping keras.ops.signbit.

Silu(*args, **kwargs)

Operation wrapping keras.ops.silu.

Sin(*args, **kwargs)

Operation wrapping keras.ops.sin.

Sinc(*args, **kwargs)

Operation wrapping keras.ops.sinc.

Sinh(*args, **kwargs)

Operation wrapping keras.ops.sinh.

Size(*args, **kwargs)

Operation wrapping keras.ops.size.

Slogdet(*args, **kwargs)

Operation wrapping keras.ops.slogdet.

SoftShrink(*args, **kwargs)

Operation wrapping keras.ops.soft_shrink.

Softmax(*args, **kwargs)

Operation wrapping keras.ops.softmax.

Softplus(*args, **kwargs)

Operation wrapping keras.ops.softplus.

Softsign(*args, **kwargs)

Operation wrapping keras.ops.softsign.

Sort(*args, **kwargs)

Operation wrapping keras.ops.sort.

SpaceToDepth(*args, **kwargs)

Operation wrapping keras.ops.space_to_depth.

SparsePlus(*args, **kwargs)

Operation wrapping keras.ops.sparse_plus.

SparseSigmoid(*args, **kwargs)

Operation wrapping keras.ops.sparse_sigmoid.

Sparsemax(*args, **kwargs)

Operation wrapping keras.ops.sparsemax.

Split(*args, **kwargs)

Operation wrapping keras.ops.split.

Sqrt(*args, **kwargs)

Operation wrapping keras.ops.sqrt.

Square(*args, **kwargs)

Operation wrapping keras.ops.square.

Squareplus(*args, **kwargs)

Operation wrapping keras.ops.squareplus.

Squeeze(*args, **kwargs)

Operation wrapping keras.ops.squeeze.

Stack(*args, **kwargs)

Operation wrapping keras.ops.stack.

Std(*args, **kwargs)

Operation wrapping keras.ops.std.

Stft(*args, **kwargs)

Operation wrapping keras.ops.stft.

Sum(*args, **kwargs)

Operation wrapping keras.ops.sum.

Svd(*args, **kwargs)

Operation wrapping keras.ops.svd.

Swapaxes(*args, **kwargs)

Operation wrapping keras.ops.swapaxes.

Swish(*args, **kwargs)

Operation wrapping keras.ops.swish.

Take(*args, **kwargs)

Operation wrapping keras.ops.take.

TakeAlongAxis(*args, **kwargs)

Operation wrapping keras.ops.take_along_axis.

Tan(*args, **kwargs)

Operation wrapping keras.ops.tan.

Tanh(*args, **kwargs)

Operation wrapping keras.ops.tanh.

TanhShrink(*args, **kwargs)

Operation wrapping keras.ops.tanh_shrink.

Threshold(*args, **kwargs)

Operation wrapping keras.ops.threshold.

Tile(*args, **kwargs)

Operation wrapping keras.ops.tile.

TopK(*args, **kwargs)

Operation wrapping keras.ops.top_k.

Trace(*args, **kwargs)

Operation wrapping keras.ops.trace.

Transpose(*args, **kwargs)

Operation wrapping keras.ops.transpose.

Tril(*args, **kwargs)

Operation wrapping keras.ops.tril.

Triu(*args, **kwargs)

Operation wrapping keras.ops.triu.

Trunc(*args, **kwargs)

Operation wrapping keras.ops.trunc.

Unfold(*args, **kwargs)

Operation wrapping keras.ops.unfold.

Unstack(*args, **kwargs)

Operation wrapping keras.ops.unstack.

Vander(*args, **kwargs)

Operation wrapping keras.ops.vander.

Var(*args, **kwargs)

Operation wrapping keras.ops.var.

View(*args, **kwargs)

Operation wrapping keras.ops.view.

ViewAsComplex(*args, **kwargs)

Operation wrapping keras.ops.view_as_complex.

ViewAsReal(*args, **kwargs)

Operation wrapping keras.ops.view_as_real.

Vsplit(*args, **kwargs)

Operation wrapping keras.ops.vsplit.

ZerosLike(*args, **kwargs)

Operation wrapping keras.ops.zeros_like.

Exceptions

MissingKerasOps(class_name, func)

class zea.ops.keras_ops.Abs(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.abs.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Absolute(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.absolute.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.AffineTransform(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.affine_transform.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.All(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.all.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Amax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.amax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Amin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.amin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Angle(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.angle.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Any(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.any.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arccos(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arccos.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arccosh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arccosh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arcsin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arcsin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arcsinh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arcsinh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arctan(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arctan.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Arctanh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.arctanh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Argmax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.argmax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Argmin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.argmin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Argpartition(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.argpartition.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Argsort(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.argsort.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Array(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.array.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ArraySplit(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.array_split.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Average(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.average.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Bartlett(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bartlett.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BatchNormalization(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.batch_normalization.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Bincount(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bincount.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseAnd(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_and.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseInvert(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_invert.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseLeftShift(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_left_shift.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseNot(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_not.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseOr(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_or.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseRightShift(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_right_shift.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BitwiseXor(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.bitwise_xor.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Blackman(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.blackman.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.BroadcastTo(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.broadcast_to.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cast(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cast.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cbrt(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cbrt.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Ceil(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ceil.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Celu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.celu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cholesky(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cholesky.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.CholeskyInverse(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cholesky_inverse.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Clip(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.clip.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Conj(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.conj.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Conjugate(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.conjugate.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ConvertToNumpy(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.convert_to_numpy.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ConvertToTensor(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.convert_to_tensor.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Copy(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.copy.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Corrcoef(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.corrcoef.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cos(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cos.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cosh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cosh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.CountNonzero(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.count_nonzero.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.CropImages(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.crop_images.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cumprod(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cumprod.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Cumsum(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.cumsum.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Deg2rad(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.deg2rad.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.DepthToSpace(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.depth_to_space.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Det(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.det.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Diag(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.diag.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Diagflat(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.diagflat.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Diagonal(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.diagonal.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Digitize(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.digitize.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Dtype(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.dtype.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Eig(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.eig.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Eigh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.eigh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ElasticTransform(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.elastic_transform.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Elu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.elu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.EmptyLike(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.empty_like.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Erf(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.erf.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Erfinv(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.erfinv.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Exp(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.exp.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Exp2(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.exp2.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ExpandDims(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.expand_dims.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Expm1(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.expm1.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ExtractPatches(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.extract_patches.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ExtractSequences(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.extract_sequences.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Fft(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.fft.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Fft2(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.fft2.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Flip(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.flip.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Fliplr(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.fliplr.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Flipud(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.flipud.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Floor(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.floor.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Fold(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.fold.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.FullLike(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.full_like.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.GaussianBlur(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.gaussian_blur.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Gelu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.gelu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.GetItem(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.get_item.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Glu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.glu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Hamming(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hamming.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Hanning(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hanning.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HardShrink(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hard_shrink.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HardSigmoid(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hard_sigmoid.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HardSilu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hard_silu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HardSwish(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hard_swish.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HardTanh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hard_tanh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Histogram(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.histogram.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Hsplit(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.hsplit.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.HsvToRgb(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.hsv_to_rgb.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.I0(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.i0.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Ifft2(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ifft2.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Imag(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.imag.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Inv(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.inv.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Irfft(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.irfft.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.IsTensor(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.is_tensor.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isfinite(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isfinite.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isinf(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isinf.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isnan(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isnan.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isneginf(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isneginf.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isposinf(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isposinf.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Isreal(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.isreal.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Istft(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.istft.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Kaiser(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.kaiser.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LayerNormalization(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.layer_normalization.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LeakyRelu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.leaky_relu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LeftShift(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.left_shift.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Log(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Log10(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log10.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Log1p(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log1p.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Log2(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log2.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LogSigmoid(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log_sigmoid.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LogSoftmax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.log_softmax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Logdet(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.logdet.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LogicalNot(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.logical_not.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Logsumexp(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.logsumexp.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.LuFactor(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.lu_factor.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Max(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.max.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Mean(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.mean.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Median(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.median.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Meshgrid(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.meshgrid.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Min(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.min.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

exception zea.ops.keras_ops.MissingKerasOps(class_name, func)[source]

Bases: ValueError

class zea.ops.keras_ops.Moments(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.moments.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Moveaxis(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.moveaxis.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.NanToNum(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nan_to_num.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanargmax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanargmax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanargmin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanargmin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nancumprod(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nancumprod.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nancumsum(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nancumsum.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanmax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanmax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanmean(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanmean.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanmedian(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanmedian.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanmin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanmin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanprod(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanprod.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanquantile(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanquantile.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanstd(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanstd.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nansum(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nansum.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nanvar(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nanvar.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Ndim(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ndim.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Negative(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.negative.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Nonzero(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.nonzero.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Norm(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.norm.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Normalize(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.normalize.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.OneHot(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.one_hot.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.OnesLike(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ones_like.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Pad(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.pad.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.PadImages(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.pad_images.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.PerspectiveTransform(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.perspective_transform.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Prod(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.prod.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Ptp(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ptp.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Qr(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.qr.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Quantile(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.quantile.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Rad2deg(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.rad2deg.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Ravel(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.ravel.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Real(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.real.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Reciprocal(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.reciprocal.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Relu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.relu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Relu6(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.relu6.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Repeat(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.repeat.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Reshape(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.reshape.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Resize(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.resize.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Rfft(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.rfft.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.RgbToGrayscale(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.rgb_to_grayscale.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.RgbToHsv(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.rgb_to_hsv.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.RightShift(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.right_shift.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.RmsNormalization(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.rms_normalization.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Roll(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.roll.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Round(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.round.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Rsqrt(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.rsqrt.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.SaturateCast(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.saturate_cast.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ScaleAndTranslate(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.image.scale_and_translate.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Selu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.selu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Shape(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.shape.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sigmoid(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sigmoid.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sign(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sign.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Signbit(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.signbit.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Silu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.silu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sin(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sin.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sinc(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sinc.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sinh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sinh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Size(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.size.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Slogdet(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.slogdet.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.SoftShrink(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.soft_shrink.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Softmax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.softmax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Softplus(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.softplus.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Softsign(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.softsign.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sort(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sort.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.SpaceToDepth(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.space_to_depth.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.SparsePlus(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sparse_plus.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.SparseSigmoid(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sparse_sigmoid.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sparsemax(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sparsemax.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Split(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.split.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sqrt(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sqrt.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Square(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.square.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Squareplus(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.squareplus.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Squeeze(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.squeeze.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Stack(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.stack.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Std(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.std.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Stft(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.stft.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Sum(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.sum.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Svd(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.svd.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Swapaxes(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.swapaxes.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Swish(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.swish.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Take(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.take.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.TakeAlongAxis(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.take_along_axis.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Tan(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.tan.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Tanh(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.tanh.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.TanhShrink(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.tanh_shrink.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Threshold(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.threshold.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Tile(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.tile.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.TopK(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.top_k.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Trace(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.trace.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Transpose(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.transpose.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Tril(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.tril.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Triu(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.triu.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Trunc(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.trunc.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Unfold(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.unfold.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Unstack(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.unstack.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Vander(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.vander.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Var(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.var.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.View(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.view.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ViewAsComplex(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.view_as_complex.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ViewAsReal(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.view_as_real.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.Vsplit(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.vsplit.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.

class zea.ops.keras_ops.ZerosLike(*args, **kwargs)[source]

Bases: Lambda

Operation wrapping keras.ops.zeros_like.

Parameters:

  • input_data_type (DataTypes or None) – Expected data type of the input tensor. Used for pipeline data-type validation; pass None to skip.

  • output_data_type (DataTypes or None) – Data type produced by this operation.

  • key (str or None) – Dict key the operation reads from (and writes to by default). Defaults to "data".

  • output_key (str or None) – Dict key the operation writes its result to. Defaults to key. Set to a different value to preserve the original input under key while producing a new key for downstream operations.

  • cache_inputs (bool) – When True, values stored via set_input_cache() are merged into every call. False means the cache is empty by default. Selective per-key caching is not supported; use set_input_cache() directly to control which keys are stored.

  • cache_outputs (bool) – Memoize outputs keyed by a hash of the merged inputs.

  • jit_compile (bool) – Wrap call() with jit() for faster execution. Disable for easier interactive debugging.

  • with_batch_dim (bool) – Whether inputs carry a leading batch dimension. Affects default axis selection in filter-type operations.

  • jit_kwargs (dict or None) – Extra keyword arguments forwarded to the JIT compiler.

  • jittable (bool) – Mark the operation as JIT-compilable. Set to False for operations that use Python control flow incompatible with tracing.

  • additional_output_keys (list of str or None) – Extra dict keys this operation may produce beyond output_key. Used for pipeline key-availability validation. Defaults to the class-level ADD_OUTPUT_KEYS list.