Capability index: GeoStack

Layers: GeoStack · GeoNumerics · GeoML · GeoModels.

Every public name in the libraries geostack builds on, each listed once under the shortest module path that exports it, with the first sentence of its docstring. Search this page before writing a loader, a model, a training step or a helper in a script or a notebook: if what you need is here, use it; if it is almost here, extend it where it lives. Regenerate with make capabilities after changing a public API (tests/test_capabilities.py checks it is current).

This page is the GeoStack layer. It imports no other layer.

Listed at pipekit 0.0.2, pipekit-array 0.0.2, pipekit-cycle 0.0.2, pipekit-evaluate 0.0.2, pipekit-experiment 0.0.2, pipekit-jax 0.0.2, pipekit-train 0.0.2, georeader-spaceml 2.4.1, geotoolz-ops 0.10.3, geotoolz-catalog 0.5.1, geotoolz-cloud 0.4.1, geotoolz-patcher 0.10.2, geotoolz-products 0.4.2, xrtoolz 0.0.5, xrtoolz-core 0.0.2, xrtoolz-einx 0.0.1, xrtoolz-grad 0.0.3, xrtoolz-reader 0.0.4, xrtoolz-sklearn 0.0.2.

Upstream: `pipekit`¶

Carrier-agnostic operator-graph framework for pipelines.

`pipekit`¶

Name	Kind	What it does
`AssertCallable`	class	Pass-through; raise `QCError` if `predicate(x)` is falsy.
`AssertDType`	class	Pass-through; raise `QCError` if `getattr(x, 'dtype')` mismatches.
`AssertHasAttribute`	class	Pass-through; raise `QCError` if attribute missing (or value mismatches).
`AssertShape`	class	Pass-through; raise `QCError` if `getattr(x, 'shape')` mismatches.
`AsyncMap`	class	Apply `op` to each item of an iterable inside an async event loop.
`BatchedMap`	class	Split an iterable into batches and call `op` on each batch as a list.
`Branch`	class	Binary conditional: run `if_true` or `if_false` based on a predicate.
`Cache`	class	Memoise an inner operator by hashing its input and config.
`Carrier`	class	Re-exported from `typing`.
`CarryState`	class	Base class for state threaded through a stateful pipeline.
`Coalesce`	class	Try each source in order; return the first output that satisfies `is_ok`.
`ConfigMixin`	class	Auto-derive `get_config()` from `__init__` parameter names.
`Const`	class	Return a fixed value, ignoring any input.
`Fanout`	class	Apply each named operator to the same input; return a dict of outputs.
`FittableTransformer`	class	Anything with sklearn-style `fit` + `transform` methods.
`Graph`	class	A symbolic operator graph with multiple inputs and outputs.
`Histogram`	class	Controller that captures simple histograms by key.
`Identity`	class	Explicit no-op.
`Input`	class	A named entry point into a `Graph`.
`Lambda`	class	Wrap an arbitrary callable as an `Operator`.
`Node`	class	A vertex in a `Graph`.
`Operator`	class	Base class for pipekit operators.
`Predictor`	class	Anything with a sklearn-style `predict` method.
`ProcessMap`	class	Apply `op` to each item of an iterable using a process pool.
`Profile`	class	Controller that records per-step wall-clock timings.
`QCError`	class	Raised when a QC assertion fails.
`Quarantine`	class	Non-raising QC: on failure log and return `sentinel`; on success pass through.
`Retry`	class	Retry an operator on transient failures with exponential backoff.
`Sequential`	class	Apply a list of operators in order, threading the output of each into the next.
`ShapeTrace`	class	Log `shape` / `dtype` of each carrier passing through.
`Signature`	class	Immutable shape + dtype record.
`Sink`	class	Apply a side-effecting write and pass the input through unchanged.
`Snapshot`	class	Controller that captures intermediate carriers by name.
`StatefulOperator`	class	Marker base for operators whose `_apply` threads carry-state.
`Switch`	class	Multi-way dispatch on `key(carrier)`.
`Tap`	class	Fire a callback on the carrier; return the carrier unchanged.
`ThreadMap`	class	Apply `op` to each item of an iterable using a thread pool.
`Try`	class	Run `primary`; on caught exceptions run `fallback`.
`allowed`	function	Return the registered class names (`"module.Class"` keys).
`check_pickleable`	function	Return the list of `forbid_in_yaml`-flagged operators inside `op`.
`complement`	function	Return a predicate that negates `predicate`.
`compose`	function	Right-to-left composition (mathematical order).
`compose_left`	function	Left-to-right composition (pipeline order).
`compute_output_signature`	function	Free-function dispatch to `op.compute_output_signature`.
`dumps`	function	JSON-encode an operator’s state record.
`juxt`	function	Return a callable that applies each operator to its input and tuples the results.
`loads`	function	Decode an operator from its JSON state record.
`loads_sandboxed`	function	Like `loads`, but reject any class not in the allowlist.
`pipe`	function	Apply `operators` left-to-right to `value`.
`register`	function	Register `cls` as loadable via `loads_sandboxed`.
`unregister`	function	Remove `cls` from the sandboxed loader’s allowlist.

Upstream: `pipekit-array`¶

Array API operators (NumPy, JAX, CuPy, PyTorch).

`pipekit_array`¶

Name	Kind	What it does
`ConcatenateAlong`	class	Concatenate a list-of-arrays along an existing axis via `xp.concat`.
`MeanScalar`	class	Reduce an array to a scalar (or per-axis means) via `xp.mean`.
`StackAlong`	class	Stack a list-of-arrays along a new axis via `xp.stack`.
`array_namespace`	function	Return the Array API namespace shared by all input arrays.

Upstream: `pipekit-cycle`¶

Time-stepping, data assimilation and observation operators.

`pipekit_cycle`¶

Name	Kind	What it does
`AdjointSpec`	constant	`pipekit_cycle.adjoints.DirectAdjoint \| pipekit_cycle.adjo...`
`AnalysisStep`	class	Combine forecast state with observations to produce analysis state.
`BFNCycle`	class	Back-and-Forth Nudging over an assimilation window.
`BacksolveAdjoint`	class	Continuous adjoint: solve the adjoint ODE backwards (O(1) memory).
`CallableForward`	class	Wrap a plain callable as a forward model.
`CallableObs`	class	Arbitrary callable observation operator.
`CompositeForward`	class	Compose multiple forward models — `F_k ∘ … ∘ F_1`.
`CompositeObs`	class	Compose multiple observation operators: `H_k ∘ … ∘ H_1`.
`CovarianceLocalizer`	class	Taper spurious long-range covariances in ensemble methods.
`Cycle`	class	Apply `step_op` for `n_steps`, threading state through each step.
`DACycle`	class	Classic forecast-analysis cycle.
`DAState`	class	State threaded through a DA cycle.
`DirectAdjoint`	class	Exact gradients by storing every step of the rollout.
`EnsembleCycle`	class	Run `Cycle` over a list of ensemble-member carriers.
`EnsembleDACycle`	class	Ensemble-based DA — EnKF, ETKF, LETKF, particle filters, …
`EnsembleInflator`	class	Inflate an analysis ensemble to counter variance underestimation.
`ErrorSubspace`	class	A propagating low-rank error-covariance factor `L` (`B ≈ L Lᵀ`).
`ForwardModel`	class	Advance a model state forward in time by `dt`.
`IdentityObs`	class	Direct measurement: `H(x) = x`.
`ImplicitAdjoint`	class	Differentiate a converged fixed point via the implicit function theorem.
`IterationState`	class	State threaded through a `Recurrence`.
`IterativeProcess`	class	An iterative ensemble inversion / sampling process (EKI, EKS, UKI).
`LatentMap`	class	An encode/decode pair between full state space and a latent space.
`LinearObs`	class	Linear observation map: `H(x) = H @ x`.
`NeuralForward`	class	Wrap a neural-emulator operator (e.g. `ModelOp`) as a `ForwardModel`.
`ObservationNoise`	class	An observation-error model that exposes covariance and sampling.
`ObservationOperator`	class	Map model state → predicted observations.
`Recurrence`	class	Apply `step_op` repeatedly until `condition_op` returns truthy.
`RecursiveCheckpointAdjoint`	class	Exact gradients with treeverse checkpointing (the recommended default).
`ReducedBasis`	class	Map a reduced control vector to a state increment (with a prior).
`ReducedOrderModel`	class	A Galerkin / POD reduced-order model: encode/decode + latent dynamics.
`SmootherCycle`	class	4D-Var-shaped windowed smoother.
`StepScheduler`	class	Step-size policy for an `IterativeProcess`.
`SupportsAdjoint`	class	A differentiable component that exposes its adjoint strategy.
`TangentLinearModel`	class	Tangent-linear and adjoint of a `ForwardModel`’s dynamics.
`TruncatedAdjoint`	class	Biased gradients: flow stops more than `k` steps back.
`WindowState`	class	State threaded through a windowed / smoother cycle.
`WindowedCycle`	class	Sliding-window time-stepping over a stream of initial carriers.

Upstream: `pipekit-evaluate`¶

Multidimensional evaluation (Unit x Lens x Stage).

`pipekit_evaluate`¶

Name	Kind	What it does
`Axis`	class	Which benchmark axis a reference is resolved along.
`Benchmark`	class	Run the `level x rung x complexity` cube and score each cell.
`BenchmarkReport`	class	The collected results of a benchmark run.
`Cell`	class	One cell of the benchmark cube.
`CellResult`	class	The scored outcome for one cube cell.
`Estimate`	class	A point and/or a posterior produced by an estimator.
`Estimator`	class	Any rung that maps observations → an :class:`Estimate`.
`Oracle`	class	The designated reference :class:`Estimator` for a task (usually rung 2).
`Prior`	class	A minimal prior over task parameters `θ`.
`ReferenceRule`	class	Resolve the reference source for a rung along a given axis.
`ReferenceSource`	class	Where a rung/step’s ground truth comes from.
`Scorer`	class	Score one prediction against one reference.
`Task`	class	One benchmark scenario: prior + simulator + obs operator + data.

Upstream: `pipekit-experiment`¶

Experiment tracking adapters and a model registry.

`pipekit_experiment`¶

Name	Kind	What it does
`ExperimentTracker`	class	Log metrics, parameters, and artifacts from a training run.
`InferenceArtifact`	class	Reproducibility snapshot for an inference pipeline.
`LocalModelRegistry`	class	Local-filesystem model registry.
`ModelRegistry`	class	Store and retrieve trained models by content hash.
`Run`	class	A single execution of a pipeline (training or inference).
`RunArtifacts`	class	Named artifacts (files, models, plots) attached to a `Run`.
`RunMetrics`	class	Per-step metric series.
`S3ModelRegistry`	class	fsspec-backed model registry.
`TrainingArtifact`	class	Reproducibility snapshot for a training run.

Upstream: `pipekit-jax`¶

JAX / Equinox carrier integration for pipekit.

`pipekit_jax`¶

Name	Kind	What it does
`DiffraxForwardModel`	lazy	Needs `diffrax`.
`JaxModelOp`	class	Wrap an `eqx.Module` as a `pipekit.Operator` with weight round-trip.
`to_diffrax_adjoint`	function	Map a pipekit adjoint spec onto a `diffrax.AbstractAdjoint`.
`truncated_scan`	function	`jax.lax.scan` with gradients truncated to the trailing `k` steps.

Upstream: `pipekit-train`¶

Training pipelines for emulators and amortized inference.

`pipekit_train`¶

Name	Kind	What it does
`CachedDataset`	class	Disk- or cloud-backed cache around any `TrainingDataset`.
`Callback`	class	Per-step / per-epoch / per-eval hooks.
`Checkpoint`	class	Save (model_op, optimizer_state) every N steps; final to a registry.
`Composite`	class	Weighted sum of `Loss` components.
`EarlyStopping`	class	Signal stop when a monitored metric hasn’t improved for N evals.
`HyperSweep`	class	Orchestrate N independent `TrainingLoop` runs.
`IterableDataset`	class	Generic escape hatch — wrap any iterable as a `TrainingDataset`.
`JSONLWriter`	class	Append-only JSON Lines metric writer.
`KL`	class	Kullback-Leibler divergence between two distributions.
`LogToExperiment`	class	Forward metrics / params / artifacts to a pipekit-experiment tracker.
`Loss`	class	Loss function — per-batch scalar called by the adapter.
`MSE`	class	Mean squared error loss.
`MetricWriter`	class	Where per-step metrics go when no tracker is attached.
`NLL`	class	Negative log-likelihood for distribution-shaped predictions.
`ParameterGrid`	class	Cartesian product over named parameter axes.
`SimulationDataset`	class	Yields `(parameters, simulator_output)` pairs.
`SweepCarryState`	class	Carry-state for `HyperSweep`.
`SweepResult`	class	One trial’s result inside a sweep.
`TrainerCarryState`	class	Carry-state for `TrainingLoop`.
`TrainingDataset`	class	Abstract base — yields `(input, target)` pairs.
`TrainingLoop`	class	Train a `model_op` on a `dataset`, producing a trained model_op.
`ValidationStep`	class	Compute validation metrics on a held-out dataset.
`XarrayWindowDataset`	lazy	Needs an optional dependency.

Upstream: `georeader-spaceml`¶

Raster reading and the GeoTensor carrier.

`georeader.geotensor`¶

Name	Kind	What it does
`GeoTensor`	class	A numpy ndarray subclass with geospatial metadata.
`get_rio_options_path`	function

`georeader.read`¶

Name	Kind	What it does
`apply_anti_aliasing`	function	Apply anti-aliasing to `data_in` assuming it will be downsampled to `resolution_dst`.
`calculate_transform_window`	function	Calculate the default transform to reproject data to dst_crs with resolution_dst_crs
`read_from_bounds`	function	Extract raster data within a geographic bounding box, with optional CRS transformation.
`read_from_center_coords`	function	Extract a rectangular chip from raster data centered on geographic coordinates.
`read_from_polygon`	function	Extract raster data within a polygon boundary, supporting complex shapes and masking.
`read_from_tile`	function	Read raster data corresponding to a Web Mercator (XYZ) tile for web mapping applications.
`read_from_window`	function	Read raster data from a specified window, with optional padding for out-of-bounds areas.
`read_reproject`	function	Reproject raster data to a different CRS, resolution, and/or extent.
`read_reproject_like`	function	Reads from `data_in` and reprojects to have the same extent and resolution than `data_like`.
`read_rpcs`	function	This function georreferences an array using the RPCs.
`read_to_crs`	function	Change the crs of data_in to dst_crs.
`resize`	function	Resample raster data to a different spatial resolution with optional anti-aliasing.
`window_from_bounds`	function	Calculate the raster window corresponding to geographic bounds.
`window_from_center_coords`	function	Calculate a raster window of specified size centered on geographic coordinates.
`window_from_polygon`	function	Calculate the raster window that contains a polygon in pixel coordinates.
`window_from_tile`	function	Calculate the raster window corresponding to a Web Mercator (XYZ) tile.

`georeader.save`¶

Name	Kind	What it does
`save_cog`	function	Save a GeoData object as a Cloud Optimized GeoTIFF (COG).
`save_tiled_geotiff`	function	Save a GeoData object as a tiled GeoTIFF file.

`georeader.rasterio_reader`¶

Name	Kind	What it does
`RasterioReader`	class	Lazy file-backed raster reader with geospatial metadata.
`get_out_shape`	function	Calculate aspect-preserving output shape.
`mask_sas_token`	function	Mask the signature of an Azure SAS token embedded in a file path.
`needs_boundless`	function	Check if a read window requires boundless mode.
`read_out_shape`	function	Read raster with resampling to target shape.

`georeader.mosaic`¶

Name	Kind	What it does
`spatial_mosaic`	function	Create a spatial mosaic by filling gaps with data from overlapping rasters.

Upstream: `geotoolz-ops`¶

Composable remote-sensing operators on GeoTensor.

`geoops`¶

Name	Kind	What it does
`ARVI`	class	Atmospherically Resistant Vegetation Index — Kaufman & Tanre 1992.
`AdaptiveWindowBackground`	class	Per-pixel local mean and diagonal variance over a square window.
`AltitudeMask`	class	Build a boolean mask from DEM elevation bounds.
`AnnotatePoints`	class	Draw circular point markers into a display GeoTensor.
`AnnotatePolygons`	class	Rasterize polygon outlines into a display GeoTensor.
`AntimeridianSplit`	class	Split a geographic swath at a longitude wrap.
`AppendIndex`	class	Run an index operator and concatenate its output back as a new band.
`ApplyAdaptiveMF`	class	Apply a matched filter against per-pixel local window statistics.
`ApplyClusterMF`	class	Apply a matched filter dispatched by per-pixel cluster label.
`ApplyColormap`	class	Map a single-band raster to a four-band RGBA GeoTensor.
`ApplyDiscreteColormap`	class	Map integer categories to a four-band RGBA GeoTensor.
`ApplyMask`	class	Apply a boolean mask to the carrier, filling True pixels.
`ApplySRF`	class	Convolve hyperspectral bands to target Gaussian-SRF multispectral bands.
`AsinhScale`	class	Inverse-hyperbolic-sine scaling.
`AtmosphericHaze`	class	Add a sampled haze term following an inverse fourth-power spectrum.
`BAIS2`	class	Burned Area Index for Sentinel-2 — Filipponi 2018.
`BAPComposite`	class	Best Available Pixel compositing from quality-score metadata.
`BBoxMask`	class	Rasterize an axis-aligned bounding box into a boolean mask.
`BSI`	class	Bare Soil Index — Rikimaru et al. 2002.
`BTFromRadiance`	class	Convert thermal at-sensor radiance to brightness temperature (Kelvin).
`BandDropout`	class	Fill each band independently with probability `p`.
`BandJitter`	class	Permute bands within explicitly configured groups.
`BandMath`	class	Evaluate a restricted arithmetic expression over named bands.
`BandRatio`	class	Simple two-band ratio `numerator / (denominator + eps)`.
`BilateralDenoise`	class	Edge-aware denoise using a range-weighted Gaussian.
`BlendMatched`	class	Weighted mean across N aligned tensors.
`BlobDOG`	class	Blob detection via Difference of Gaussian.
`BlobDoH`	class	Blob detection via Determinant of Hessian.
`BlobLoG`	class	Blob detection via Laplacian of Gaussian.
`BowtieCorrection`	class	Remove whiskbroom bowtie overlap by resampling each scan along track.
`BrightnessJitter`	class	Scale reflectance values by a sampled scalar or per-band factor.
`BufferMask`	class	Radially expand True pixels in a mask by a Euclidean buffer.
`CHWtoHWC`	class	Reorder a `(..., C, H, W)` cube to channels-last `(..., H, W, C)`.
`CIRI`	class	Cirrus Reflectance Index — Sentinel-2 B10 passthrough.
`CLAHE`	class	Contrast-Limited Adaptive Histogram Equalization.
`Canny`	class	Canny edge detection returning a boolean edge map.
`ChanVese`	class	Active-contour segmentation via :func:`skimage.segmentation.chan_vese`.
`ClayMinerals`	class	Clay Minerals ratio — Sabins 1999 / Crowley et al. 1989.
`CleanMask`	class	Remove small objects, fill small holes, then close the mask.
`CloseMask`	class	Close (dilate then erode) a boolean mask.
`CloudFreeComposite`	class	Per-pixel mean over frames where the cloud mask is false.
`CloudSEN12`	class	Placeholder for the optional ML-based CloudSEN12 detector.
`ColumnEnhancement`	class	Convenience trace-gas column-enhancement matched-filter composite.
`ColumnToMass`	class	Convert column enhancement units (ppm m / mol m^-2 / kg m^-2).
`CombineMasks`	class	Combine boolean masks on one pixel grid with `or`, `and` or `xor`.
`Composite`	class	Build a multi-band composite by arbitrary band reference.
`ComputeSZA`	class	Compute the solar zenith angle (degrees) for a location and UTC datetime.
`ContinuumRemoval`	class	Hull-quotient continuum removal along the band axis.
`ContrastJitter`	class	Scale deviations from the per-band spatial mean.
`CornerHarris`	class	Harris corner response plus peak selection as point features.
`CountryMask`	class	Rasterize Natural Earth country polygons selected by ISO A3 code.
`CropTo`	class	Crop a `GeoTensor` to a target spatial shape.
`CropToBounds`	class	Crop a `GeoTensor` to a geographic bounding box.
`CrossSectionalFlux`	class	Estimate cross-sectional flux at downwind transects.
`CutMix`	class	Paste a random rectangle from a pool donor on the input’s pixel grid.
`DNToRadiance`	class	Convert DN to at-sensor radiance.
`DNToReflectance`	class	Convert DN to TOA / surface reflectance via a linear affine decode.
`DOS1`	class	Apply a Chavez (1988) Dark-Object Subtraction (DOS1) approximation.
`DecodeBitmask`	class	Unpack a QA bitmask into named boolean mask layers.
`DenoisePCA`	class	Reconstruct a multi-band raster from its top PCA components.
`DespeckleFrost`	class	Frost (1982) adaptive exponential-kernel speckle filter.
`DespeckleLee`	class	Lee (1980) local-statistics speckle filter.
`DespeckleRefinedLee`	class	Refined-Lee speckle filter approximation.
`DestripeColumn`	class	Subtract per-column (or per-row) offsets from striped imagery.
`DetectionThreshold`	class	Gaussian false-alarm score threshold.
`DilateMask`	class	Dilate a boolean mask over its trailing spatial axes.
`DistanceMask`	class	Mask pixels by their distance from a geometry (True = drop).
`EVI`	class	Enhanced Vegetation Index — Huete et al. 2002.
`EVI2`	class	Two-band Enhanced Vegetation Index — Jiang et al. 2008.
`EarthSunDistanceCorrection`	class	Compute the Earth–Sun distance `d` (in AU) for an acquisition date.
`Einx`	class	Apply any einx operation to the carrier, geo-aware.
`ErodeMask`	class	Erode a boolean mask over its trailing spatial axes.
`EstimateCovEmpirical`	class	Estimate an empirical covariance operator.
`EstimateCovLowRank`	class	Estimate a low-rank-plus-Tikhonov covariance operator.
`EstimateCovShrunk`	class	Estimate a diagonal-target shrinkage covariance operator.
`EstimateMean`	class	Estimate a spectral background mean vector from a cube.
`ExpandLabels`	class	Grow label regions by a fixed pixel distance without overlap.
`FalseColor`	class	Build a NIR-red-green false-colour composite.
`Felzenszwalb`	class	Graph-based segmentation via :func:`skimage.segmentation.felzenszwalb`.
`FindContours`	class	Extract iso-value contours as LineString geometries.
`GCI`	class	Green Chlorophyll Index — Gitelson et al. 2003.
`GMMClusterBackground`	class	Estimate clustered background statistics with a NumPy GMM backend.
`Gamma`	class	Power-law gamma correction.
`GammaCorrect`	class	Apply display gamma correction.
`GapFillIDW`	class	Fill NaNs with inverse-distance weighted finite neighbours.
`GapFillInpaintBiharmonic`	class	Smoothly inpaint NaN gaps with a biharmonic-style fill.
`GapFillLaplacian`	class	Fill NaNs by solving a discrete Laplace equation.
`GapFillNearest`	class	Fill NaNs from the nearest finite neighbour.
`GaussianDenoise`	class	Gaussian smoother over the trailing two spatial axes.
`GaussianNoise`	class	Add zero-mean Gaussian sensor noise.
`GeoTensorEstimator`	class	Marshall a scikit-learn estimator to and from GeoTensor-shaped data.
`Georeference`	class	Georeference swath data using a `georeader` GLT `GeoTensor`.
`GeostationaryParallaxCorrect`	class	Shift pixels from apparent geostationary view position to nadir.
`HOG`	class	Histogram of Oriented Gradients descriptor.
`HWCtoCHW`	class	Reorder a channels-last `(..., H, W, C)` array to `(..., C, H, W)`.
`Hillshade`	class	Compute a single-band `uint8` hillshade from a DEM.
`HistogramMatch`	class	Match a GeoTensor histogram to a reference GeoTensor.
`HistogramStretch`	class	Per-band percentile stretch into `out_range` for visualisation.
`HoughCircles`	class	Detect circles with the circular Hough transform.
`HoughLines`	class	Detect prominent straight lines with the Hough transform.
`IMEEstimate`	class	Integrated Mass Enhancement (IME) flux estimator.
`IntegratedIrradiance`	class	Compute band-integrated TOA solar irradiance from an SRF table.
`InvertMask`	class	Invert a boolean mask element-wise.
`IronOxide`	class	Iron Oxide ratio — Segal 1982 / Sabins 1999.
`LabelConnectedComponents`	class	Convert a binary mask into an int32 connected-component label map.
`LandMask`	class	Rasterize Natural Earth land polygons into a boolean mask.
`LandsatQA_PIXEL`	class	Landsat Collection-2 QA_PIXEL mask preset.
`LinearTargetFromObs`	class	Finite-difference tangent-linear target from an observation model.
`LoadFromEE`	class	Load an Earth Engine image into a :class:`GeoTensor`.
`LoadFromSTAC`	class	Load a raster asset from a STAC item by asset key.
`LogScale`	class	Logarithmic scaling with an epsilon offset for zeros.
`MNDWI`	class	Modified Normalized Difference Water Index — Xu 2006.
`MNF`	class	Minimum Noise Fraction transform (Green et al. 1988), with its inverse.
`MODISStateQA`	class	MODIS `state_1km` (or `state_500m`) QA mask preset.
`MarkBoundaries`	class	Overlay segmentation boundaries on an image for visual inspection.
`MaskCirrus`	class	Return True where a QA band marks cirrus pixels.
`MaskCloudShadow`	class	Return True where a QA band marks cloud-shadow pixels.
`MaskClouds`	class	Return True where a QA band marks cloud-contaminated pixels.
`MaskInvalid`	class	Mark invalid (nodata) pixels: True = invalid, drop it.
`MaskNMS`	class	Mask-IoU non-maximum suppression over a stack of instance masks.
`MaskNoData`	class	Return True where pixels are no-data by QA value or carrier fill value.
`MaskSaturated`	class	Return True where pixels are at or above a saturation value.
`MaskSnow`	class	Return True where a QA band marks snow or ice pixels.
`MaskWater`	class	Return True where a QA band marks water pixels.
`MatchedFilter`	class	Per-pixel matched-filter score map.
`MatchedFilterPixel`	class	Single-pixel matched-filter score.
`MatchedFilterSNR`	class	Theoretical detection SNR `amplitude * sqrt(t^T Sigma^-1 t)`.
`MaxNDVIComposite`	class	Pick the frame with maximum NDVI per pixel and return its band values.
`MedianComposite`	class	Per-pixel median across a stack of co-registered GeoTensors.
`MedianDenoise`	class	Median filter over the trailing two spatial axes.
`MergeNearbyInstances`	class	Merge instance labels whose bboxes are close and partially overlap.
`MinCloudComposite`	class	Pick each pixel from the scene with the lowest global cloud coverage.
`MinMaxScaler`	class	Per-band min-max scaling into an arbitrary output range.
`MinMaxStretch`	class	Linear contrast stretch into `[0, 1]`.
`ModelOp`	class	Wrap any callable as an Operator.
`MomentMatching`	class	Destripe by matching every column’s mean and std to a reference.
`Mosaic`	class	Mosaic multiple `GeoTensor`s onto a single grid.
`MultiscaleBasicFeatures`	class	General-purpose multiscale feature stack.
`NBR`	class	Normalized Burn Ratio — Key & Benson 2006.
`NBR2`	class	Normalized Burn Ratio 2 — USGS Landsat product.
`NDBI`	class	Normalized Difference Built-up Index — Zha et al. 2003.
`NDMI`	class	Normalized Difference Moisture Index — Gao 1996.
`NDSI`	class	Normalized Difference Snow Index — Hall et al. 1995.
`NDVI`	class	Normalized Difference Vegetation Index — Rouse et al. 1974.
`NDWI`	class	Normalized Difference Water Index — McFeeters 1996.
`NLMeans`	class	Lightweight non-local-means-style denoiser.
`NonlinearTargetFromObs`	class	Finite-amplitude target from a callable observation model.
`Normalize`	class	Fixed-stats per-band z-score normaliser.
`NormalizedDifference`	class	Generic normalized-difference index `(a - b) / (a + b + eps)`.
`OceanMask`	class	Rasterize Natural Earth ocean polygons into a boolean mask.
`OmniCloudMask`	class	Placeholder for the optional ML-based OmniCloudMask detector.
`OpenMask`	class	Open (erode then dilate) a boolean mask.
`OpticalFlowILK`	class	Dense per-pixel displacement via iterative Lucas-Kanade optical flow.
`OpticalFlowTVL1`	class	Dense per-pixel displacement via TV-L1 optical flow.
`OutlierMask`	class	Flag global outliers with a MAD or z-score threshold.
`Overlay`	class	Blend background and foreground GeoTensors on the same grid.
`PadTo`	class	Pad a `GeoTensor` up to a target spatial shape.
`PeakLocalMax`	class	Find local maxima and return point features with pixel scores.
`PerBandReduce`	class	Reduce each band over its spatial extent to one scalar, ignoring nodata.
`PerBandStats`	class	Compute NaN-aware per-band statistics and cache them.
`PercentileClip`	class	Per-band robust contrast stretch using percentile thresholds.
`PhaseAlign`	class	Sub-pixel image registration via phase cross-correlation.
`PixelwiseGMM`	class	Gaussian mixture convenience operator.
`PixelwiseIPCA`	class	Streaming IncrementalPCA convenience operator.
`PixelwiseIsolationForest`	class	Pixel-wise IsolationForest anomaly-score convenience operator.
`PixelwiseIterativeImputer`	class	Pixel-wise IterativeImputer convenience operator.
`PixelwiseKMeans`	class	Pixel-wise KMeans label convenience operator.
`PixelwiseKNNImputer`	class	Pixel-wise KNNImputer convenience operator.
`PixelwiseLocalOutlierFactor`	class	Pixel-wise LocalOutlierFactor convenience operator.
`PixelwiseMiniBatchKMeans`	class	Streaming MiniBatchKMeans convenience operator.
`PixelwiseNMF`	class	Pixel-wise NMF convenience operator.
`PixelwiseOneClassSVM`	class	Pixel-wise OneClassSVM anomaly-score convenience operator.
`PixelwisePCA`	class	Pixel-wise PCA convenience operator.
`PlumeColumnStats`	class	Per-instance in-mask vs background statistics of a column field.
`PlumeContours`	class	Connected-component labelling of a plume mask.
`PlumeFootprint`	class	Vectorize a plume mask into polygons with per-plume metadata.
`PlumeMask`	class	Binary plume mask from a single-band score or enhancement map.
`PlumeQNDFeatures`	class	Per-instance Quantile-Normality-Deviation features for plume QA.
`PlumeShapeFilter`	class	Drop instance labels that fail a fiber-shape morphology gate.
`PointCloudToRaster`	class	Rasterize a point cloud onto a target grid.
`PointsToRaster`	class	Bin a vector cube of point measurements onto a target raster grid.
`PolygonMask`	class	Rasterize a shapely geometry or GeoDataFrame into a boolean mask.
`PowerScale`	class	Non-negative power scaling.
`ProfileLine`	class	Sample values along a line between two pixel coordinates.
`Quickshift`	class	Mode-seeking superpixels via :func:`skimage.segmentation.quickshift`.
`RANSAC`	class	Robust model fitting via :func:`skimage.measure.ransac`.
`RGBRecipe`	class	Build a display RGB from three band expressions with fixed stretches.
`RadianceToDN`	class	Convert at-sensor radiance back to raw DN — inverse of `DNToRadiance`.
`RadianceToReflectance`	class	Convert at-sensor radiance to TOA reflectance with solar geometry.
`RandomCrop`	class	Randomly crop a spatial window and update the transform.
`RandomFlip`	class	Randomly flip a GeoTensor horizontally and/or vertically.
`RandomRotate90`	class	Randomly rotate by 90, 180, or 270 degrees.
`RandomShift`	class	Randomly shift a fixed-size spatial window in pixel units.
`RandomWalker`	class	Seeded segmentation via :func:`skimage.segmentation.random_walker`.
`RasterToPointCloud`	class	Sample raster values onto each node of a point cloud.
`RasterToPoints`	class	Extract raster values at a vector cube of points.
`RasterToRasterLike`	class	Align one raster onto another’s CRS + grid in a single op.
`Rasterize`	class	Rasterize geometries onto the input `GeoTensor`’s grid.
`RasterizeLike`	class	Rasterize geometries onto a stored reference grid.
`ReadBounds`	class	Read all pixels intersecting a geographic bounding box.
`ReadCenterCoords`	class	Read a fixed-shape window centred on map coordinates.
`ReadHDF`	class	Read a dataset from an HDF4/HDF-EOS or HDF5 file, local or remote.
`ReadNetCDF`	class	Read a variable from a NetCDF-3 / NetCDF-4 file, local or remote.
`ReadPolygon`	class	Read all pixels intersecting a polygon.
`ReadReprojectLike`	class	Read a source and reproject it onto another raster’s grid.
`ReadTile`	class	Read a single web-map tile and reproject it to a target CRS.
`ReadToCRS`	class	Read a source and reproject it to a target CRS.
`ReadWindow`	class	Read a pixel-space rectangular window from a raster source.
`ReflectanceToRadiance`	class	Convert TOA reflectance back to at-sensor radiance — inverse of `RadianceToReflectance`.
`RegionProps`	class	Extract a GeoDataFrame of per-region statistics from a label map.
`RemoveSmallHoles`	class	Fill enclosed False components up to `max_hole_area_px` pixels.
`RemoveSmallObjects`	class	Remove connected True components smaller than `min_area_px` pixels.
`ReorderBands`	class	Reorder bands by integer index or band name.
`ReplaceOutliers`	class	Replace outliers with median, NaN, or nearest-neighbour interpolation.
`Reproject`	class	Reproject a `GeoTensor` to a destination CRS (and optional resolution).
`ReprojectLike`	class	Reproject a `GeoTensor` onto another `GeoTensor`’s CRS + grid.
`Resample`	class	Resample a `GeoTensor` to a target spatial resolution.
`ResampleLike`	class	Resample a `GeoTensor` onto another `GeoTensor`’s spatial grid.
`Resize`	class	Resize a `GeoTensor` to a target spatial shape.
`RobustScaler`	class	Per-band median / IQR scaling (robust to outliers).
`S2Cloudless`	class	Placeholder for the optional ML-based s2cloudless mask.
`S2QA60`	class	Sentinel-2 L1C QA60 cloud + cirrus mask preset.
`S2SCL`	class	Sentinel-2 L2A SCL preset that masks the requested `targets` classes.
`SAVI`	class	Soil-Adjusted Vegetation Index — Huete 1988.
`SBMP`	class	Single-band multi-pass Sentinel-2 SWIR ratio retrieval.
`SCL`	class	Sentinel-2 L2A Scene Classification Layer class IDs.
`SCL_CLOUDS`	constant	`frozenset({<SCL.CLOUD_MEDIUM_PROBABILITY: 8>, <SCL.CLOUD_...`
`SCL_CLOUDS_AND_INVALID`	constant	`frozenset({<SCL.NO_DATA: 0>, <SCL.SATURATED_OR_DEFECTIVE:...`
`SCL_INVALID`	constant	`frozenset({<SCL.NO_DATA: 0>, <SCL.SATURATED_OR_DEFECTIVE:...`
`SCL_LAND`	constant	`frozenset({<SCL.VEGETATION: 4>, <SCL.NOT_VEGETATED: 5>})`
`SCL_WATER`	constant	`frozenset({<SCL.WATER: 6>})`
`SENSOR_QA_REGISTRY`	constant	`{'s2_qa60': {'cloud': {'bits': (10,)}, 'cirrus': {'bits':...`
`SLIC`	class	SLIC superpixels via :func:`skimage.segmentation.slic`.
`SWIRComposite`	class	Build a SWIR2-NIR-red composite.
`SegmentStitch`	class	Assemble indexed sensor segments into one contiguous `GeoTensor`.
`SelectBands`	class	Select bands by integer index or band name along the band axis.
`ShadedRelief`	class	Apply an elevation colormap and modulate RGB with hillshade.
`ShannonEntropy`	class	Compute Shannon entropy of the input image.
`SimpleAtmosphericCorrection`	class	Dispatch simple BOA approximations; currently supports `method='dos1'`.
`SimulatedClouds`	class	Overlay a smooth synthetic cloud field onto reflectance imagery.
`SkeletonLength`	class	Longest path through the skeleton of a binary mask.
`SklearnOp`	class	Universal adapter for scikit-learn-compatible estimators on GeoTensors.
`SlidingWindow`	class	Split a `GeoTensor` into overlapping sliding-window tiles.
`SlopeMask`	class	Build a boolean mask from DEM slope bounds in degrees.
`SpatialPool`	class	Downsample the spatial grid by integer factors, updating the transform.
`SpeckleNoise`	class	Apply multiplicative Gaussian speckle noise, useful for SAR.
`SpectralBinning`	class	Aggregate source bands into wavelength-centered bins.
`SpectralSmoothing`	class	Smooth spectra along the band axis.
`SplitBands`	class	Split a multiband GeoTensor into one single-band GeoTensor per band.
`StackBands`	class	Concatenate carriers along the band axis.
`StackMatched`	class	Concatenate aligned tensors along the band axis.
`StandardScaler`	class	Per-band z-score normalisation.
`Stitch`	class	Stitch georeferenced tiles back into one `GeoTensor`.
`StreamingBackground`	class	Aggregate mean/covariance across an iterable of cubes.
`StretchToUint8`	class	Percentile-stretch display data to `uint8`.
`StructureTensor`	class	Local structure-tensor eigenvalue stack.
`SunAngleJitter`	class	Rescale TOA reflectance for a simulated solar-zenith change.
`Threshold`	class	Binary mask of the pixels strictly above a global threshold.
`Tile`	class	Split a `GeoTensor` into spatial tiles.
`ToFloat32`	class	Cast the carrier’s values to `float32`.
`TrueColor`	class	Build an RGB composite from explicit red, green, and blue band refs.
`ValidateMFInputs`	class	Fail-fast pass-through validator for matched-filter inputs.
`VectorToRasterAgg`	class	Rasterize a vector layer with an aggregation policy for overlaps.
`Vectorize`	class	Vectorize non-zero regions of a mask `GeoTensor` into polygons.
`Watershed`	class	Watershed segmentation via :func:`skimage.segmentation.watershed`.
`WindAdvectionCone`	class	Mask likely downwind plume extent from a source and wind vector.
`WriteCOG`	class	Write a :class:`GeoTensor` as a validated Cloud-Optimized GeoTIFF.
`WriteGeoTIFF`	class	Write a :class:`GeoTensor` as a tiled GeoTIFF (no overviews).
`WriteZarr`	class	Write a :class:`GeoTensor` to a Zarr store with spatial metadata.
`ZeroOne`	class	Scale the current scene extent into `[0, 1]`.
`dNBR`	class	Differenced Normalized Burn Ratio — Key & Benson 2006.
`kNDVI`	class	Kernel NDVI — Camps-Valls et al. 2021.

Upstream: `geotoolz-catalog`¶

Spatiotemporal catalog over geospatial files.

`geocatalog`¶

Name	Kind	What it does
`GeoCatalog`	class	A queryable spatiotemporal index over geospatial files.
`GeoSlice`	class	A bounded request for data — produced by samplers, consumed by loaders.
`intersect`	function	Cross-catalog AND — rows from `left` paired with overlapping rows in `right`.
`open_catalog`	function	Open a GeoParquet artifact as a `GeoCatalog`.
`query`	function	Filter `catalog` by space + time and return the matching rows.
`union`	function	Cross-catalog OR — concatenate `left` and `right` into one catalog.

Upstream: `geotoolz-cloud`¶

Cloud object storage and batched async COG reads.

Upstream: `geotoolz-patcher`¶

Four-axis patcher: split, run per patch, stitch.

`geopatcher`¶

Name	Kind	What it does
`AsyncField`	class	Async mirror of `Field` — concurrent `select` over remote tiles.
`AsyncSpatialPatcher`	class	Async mirror of `SpatialPatcher` over an `AsyncField`.
`Domain`	class	Metadata view of a field — bounds, CRS, no I/O.
`Field`	class	A substrate the Patcher reads patches out of.
`Patch`	class	A single patch produced by a `SpatialPatcher`.
`RasterField`	class	Wrap a sync `GeoData` (a `RasterioReader`, a `GeoTensor`, …) as a `Field`.
`SpatialPatcher`	class	The four-axis spatial Patcher.
`SpatioTemporalPatch`	class	A patch carrying both a spatial and a temporal anchor.
`SpatioTemporalPatcher`	class	Composition of a spatial and a temporal Patcher.
`TemporalPatch`	class	A single patch produced by a `TemporalPatcher`.
`TemporalPatcher`	class	Four-axis temporal Patcher.

Upstream: `geotoolz-products`¶

Readers for Earth-observation data products.

`geoproducts`¶

Name	Kind	What it does
`ProductReader`	class	ABC for georeader-compatible product readers.
`stack`	function	Stack readers’ bands on the grid of `readers[like]`.

Upstream: `xrtoolz`¶

Composable operators for Earth System Data Cubes.

`xrtoolz`¶

Name	Kind	What it does
`ApplyToEach`	class	Re-instantiate a prototype operator per value and merge results.
`Augment`	class	Run an inner `Operator` and merge its output back into the input.
`Operator`	class	Re-exported from `xrcore`.
`Signature`	class	Re-exported from `xrcore`.

Upstream: `xrtoolz-core`¶

xarray-aware Operator base and Signature descriptor.

Upstream: `xrtoolz-einx`¶

Labeled named-tensor algebra bridging xarray and einx.

`xreinx`¶

Name	Kind	What it does
`BatchMatmul`	class	Batched matrix multiply: contract `dim` within shared batch dims.
`CoordMismatch`	class	Shared dims have unequal coords and `align=False`.
`Einsum`	class	Named-tensor Einstein summation over one or more DataArrays.
`EinxBridgeError`	class	Base for all :mod:`xreinx` errors.
`Matmul`	class	Contract two DataArrays over a shared dimension (matrix multiply).
`Outer`	class	Outer product of two DataArrays over their disjoint dims.
`PatternError`	class	Pattern is malformed or references dims not on the inputs.
`Rearrange`	class	Reshape / transpose a DataArray by a named-tensor pattern.
`Reduce`	class	Reduce a DataArray over named axes with a chosen reduction.
`Repeat`	class	Broadcast / tile a DataArray along new named axes.
`batch_matmul`	function	`matmul` along `dim` broadcast over shared `batch_dims`.
`einsum`	function	Labeled einx einsum (contraction) over named dims.
`matmul`	function	Contract `a` and `b` along the shared `dim`.
`outer`	function	Outer product.
`pack_dataset`	function	Stack named variables of a Dataset along a new dim.
`rearrange`	function	Labeled einx rearrange (reshape / transpose / merge / split).
`reduce`	function	Reduce over named axes (axes on the input slot, absent from output).
`repeat`	function	Broadcast / tile along new named axes.
`unpack_dataset`	function	Split a ‘variable’-style dim back into a Dataset.

Upstream: `xrtoolz-grad`¶

Finite-difference calculus on labeled xarray grids.

`xrgrad`¶

Name	Kind	What it does
`EARTH_RADIUS`	constant	`6371000.0`
`GRAVITY`	constant	`9.80665`
`OMEGA`	constant	`7.2921159e-05`
`curl`	function	2-D scalar curl `∂v/∂x − ∂u/∂y` (vertical component of `∇×F`).
`divergence`	function	Divergence `∇·F` of a vector field stored as Dataset variables.
`gradient`	function	Gradient `∇da` under the given geometry.
`grid_metrics_from_coords`	function	Derive `(volume_metrics, face_metrics)` from a Dataset’s coords.
`laplacian`	function	Laplacian `Δf = ∇·∇f`.
`partial`	function	Partial derivative `∂ᵒʳᵈᵉʳda/∂<dim>ᵒʳᵈᵉʳ` under the given geometry.

Upstream: `xrtoolz-reader`¶

Authenticated archive readers (CMEMS, CDS, AEMET).

`xrreader`¶

Name	Kind	What it does
`AEMETCredentials`	class	API key for AEMET OpenData.
`AemetArchive`	class	Append-only local mirror of AEMET station observations.
`AemetAuthError`	class	Missing or rejected API key.
`AemetError`	class	Any AEMET-level failure (auth, rate limit, bad envelope, 5xx).
`AemetRateLimitError`	class	Quota exhausted even after backoff.
`AemetSource`	class	Adapter for AEMET OpenData observation endpoints.
`ArchiveCoverage`	class	Per-station span of what the archive already holds for a preset.
`BBox`	class	Axis-aligned lon/lat bounding box.
`CATALOG`	constant	`{'glorys12.daily': CatalogEntry(source='cmems', dataset_i...`
`CDSCredentials`	class	URL / key pair for the Climate Data Store.
`CDSFormProfile`	class	Shape of a CDS `retrieve()` form for a dataset family.
`CDSInsituArchive`	class	Append-only local mirror of a CDS in-situ preset.
`CDSSource`	class	Adapter around the `cdsapi` Python client.
`CMEMSCredentials`	class	Username / password pair for Copernicus Marine.
`CMEMSSource`	class	Adapter around the `copernicusmarine` Python client.
`CatalogEntry`	class	Short-name-to-source mapping with optional request defaults.
`DataSource`	class	Abstract interface implemented by every adapter.
`DatasetInfo`	class	Static description of a remote dataset.
`DatasetKind`	class	Discriminator for how a dataset is laid out once materialized.
`DepthRange`	class	Continuous depth window (positive-down, metres).
`INSITU`	constant	`CDSFormProfile(family='insitu-land', format_key='data_for...`
`Issue`	class	Single validation finding.
`LocalL2Source`	class	Open satellite L2 product files from the local filesystem.
`PressureLevels`	class	Discrete pressure levels in hPa (CDS-style).
`REANALYSIS`	constant	`CDSFormProfile(family='reanalysis', format_key='format', ...`
`REGISTRY`	constant	`{'sst': Variable(name='sst', standard_name='sea_surface_t...`
`Request`	class	Canonical request payload consumed by data adapters.
`Severity`	class	Two-level severity for validation issues.
`Station`	class	One fixed observation station.
`StationCollection`	class	Immutable collection of :class:`Station` with set-like filters.
`TimeRange`	class	Inclusive time window, optionally with a target sampling frequency.
`ValidationReport`	class	Aggregate validation result with convenience accessors.
`Variable`	class	CF-aware variable descriptor used across adapters and validators.
`all_entries`	function	Return a shallow copy of the unified catalog.
`apply_cf_attrs`	function	Stamp CF attributes from `variable` onto `da`.
`describe`	function	Look up a :class:`DatasetInfo` by short name.
`load_aemet`	function	Load AEMET credentials, or `None` if none could be found.
`load_cds`	function	Load CDS credentials, or `None` if none could be found.
`load_cmems`	function	Load CMEMS credentials, or `None` if none could be found.
`open_emit_ch4_l2b`	function	Open an EMIT L2B CH4 enhancement scene as a flat CF Dataset.
`open_ghgsat_ch4_l2`	function	Open a GHGSat per-plume NetCDF as a flat CF Dataset.
`open_tropomi_ch4_l2`	function	Flatten a TROPOMI CH4 L2 granule into one CF Dataset.
`register`	function	Add `variable` to :data:`REGISTRY` (keyed on :attr:`Variable.name`).
`resolve`	function	Return a :class:`Variable` given an instance or its canonical name.
`subset_bbox`	function	Keep points whose `(lon, lat)` fall inside a bounding box.
`subset_time`	function	Restrict a dataset to the time range `[time_min, time_max]`.
`subset_where`	function	Keep points where `variable` is in `[min_val, max_val]`.
`validate_dataset`	function	Validate multiple variables in a dataset.
`validate_variable`	function	Validate a single :class:`xarray.DataArray` against a :class:`Variable`.

Upstream: `xrtoolz-sklearn`¶

scikit-learn to xarray bridge (XarrayEstimator, SklearnOp).

`xrsklearn`¶

Name	Kind	What it does
`MissingKind`	function	Re-exported from `typing`.
`NanPolicy`	function	Re-exported from `typing`.
`SklearnOp`	class	Layer-1 operator that delegates sklearn marshalling to `XarrayEstimator`.
`TreeMode`	function	Re-exported from `typing`.
`XarrayEstimator`	class	Wrap an sklearn estimator so it operates on xarray inputs.

Capability index: GeoStack

Upstream: pipekit¶

pipekit¶

Upstream: pipekit-array¶

pipekit_array¶

Upstream: pipekit-cycle¶

pipekit_cycle¶

Upstream: pipekit-evaluate¶

pipekit_evaluate¶

Upstream: pipekit-experiment¶

pipekit_experiment¶

Upstream: pipekit-jax¶

pipekit_jax¶

Upstream: pipekit-train¶

pipekit_train¶

Upstream: georeader-spaceml¶

georeader.geotensor¶

georeader.read¶

georeader.save¶

georeader.rasterio_reader¶

georeader.mosaic¶

Upstream: geotoolz-ops¶

geoops¶

Upstream: geotoolz-catalog¶

geocatalog¶

Upstream: geotoolz-cloud¶

Upstream: geotoolz-patcher¶

geopatcher¶

Upstream: geotoolz-products¶

geoproducts¶

Upstream: xrtoolz¶

xrtoolz¶

Upstream: xrtoolz-core¶

Upstream: xrtoolz-einx¶

xreinx¶

Upstream: xrtoolz-grad¶

xrgrad¶

Upstream: xrtoolz-reader¶

xrreader¶

Upstream: xrtoolz-sklearn¶

xrsklearn¶