kd.metrics#

Metrics.

Symbols#


kd.metrics.Accuracy	Classification Accuracy.
kd.metrics.Ari	Adjusted Rand Index (ARI) computed from predictions and labels.
kd.metrics.AutoState	Flexible base class for conveniently defining custom states.
kd.metrics.AverageState	Computes the average of a scalar or a batch of tensors.
kd.metrics.BinaryAccuracy	Classification Accuracy for Binary classification tasks.
kd.metrics.CollectFirstState	Get the first outputs (possibly) across multiple steps (no reducing).
kd.metrics.CollectingState	Accumulate outputs across multiple steps (without reducing).
kd.metrics.EmptyState	Empty state.
kd.metrics.LpipsVgg	VGG LPIPS.
kd.metrics.Metric	Base class for metrics.
kd.metrics.NoopMetric	Metric that does nothing. Can be used in sweeps to remove a metric.
kd.metrics.Norm	Wraps jnp.linalg.norm to compute the average norm for given tensors.
kd.metrics.Precision1	Precision@1 for multilabel classification.
kd.metrics.Psnr	PSNR.
kd.metrics.RocAuc	Area Under the Receiver Operating Characteristic Curve (ROC AUC).
kd.metrics.SingleDimension	Returns a single chosen dimension of the tensor.
kd.metrics.SkipIfMissing	Skip this metric if any of the keys are missing.
kd.metrics.Ssim	Structural similarity (SSIM).
kd.metrics.State	Base metric state class.
kd.metrics.Std	Compute the standard deviation for float values.
kd.metrics.TreeMap	Maps an inner metric to a pytree and returns a pytree of results.
kd.metrics.TreeReduce	Applies a metric to a pytree and returns the aggregated result.


kd.metrics.concat_field	Defines a AutoState data-field that is merged by concatenation.
kd.metrics.static_field	Define an AutoState static field.
kd.metrics.sum_field	Define an AutoState data-field that is merged by summation (a + b).
kd.metrics.truncate_field	Defines a AutoState data-field that is merged by truncation.