# kd.evals

[[[Source]]](https://github.com/google-research/kauldron/tree/main/kauldron/evals/__init__.py)

```{eval-rst}
.. automodule:: kauldron.evals
  :no-members:
```

## Symbols


### Class

|  |  |
--- | ---
[kd.evals.CollectionKeys](CollectionKeys) | Names of the metrics/summaries/losses (displayed in flatboard).
[kd.evals.Evaluator](Evaluator) | Evaluator running `num_batches` times.
[kd.evals.EvaluatorBase](EvaluatorBase) | Base class for inline evaluators.
[kd.evals.EveryNSteps](EveryNSteps) | Run eval every N train steps.
[kd.evals.FewShotEvaluator](FewShotEvaluator) | FewShotEvaluator running closed-form few-shot classification.
[kd.evals.Once](Once) | Run eval only after the `XX` train steps.
[kd.evals.RunStrategy](RunStrategy) | Base class for info on how to run the evaluation.
[kd.evals.StandaloneEveryCheckpoint](StandaloneEveryCheckpoint) | Run eval continuously everytime a new checkpoint is found.
[kd.evals.StandaloneLastCheckpoint](StandaloneLastCheckpoint) | Run eval only after the last checkpoint, after train has completed.

```{toctree}
:hidden:

CollectionKeys
Evaluator
EvaluatorBase
EveryNSteps
FewShotEvaluator
Once
RunStrategy
StandaloneEveryCheckpoint
StandaloneLastCheckpoint
```