mspruc opened a new pull request, #765:
URL: https://github.com/apache/wayang/pull/765
Wayang currently is based on manually tuned cost models, this can be time
consuming to manually tune therefore we implement
- `DefaultPointwiseCost` which is called using Wayangs native
`EstimatableCostFactory`, you just need to provide it to Wayangs `Configuration`
- `cost_model.onnx` a pretrained sample model, based off a tree
convolutional neural network for cost prediction, we also allow you to manually
set a model with `config.setProperty("wayang.ml.model.file", modelPath);`
- Introduces `TreeNode`, `TreeEncoder`, `TreeDecoder`, `OneHotEncoder`,
`OneHotMappings` encoding schemes which allows capturing deep relationships
within a plan
- Integrates `cost_model.onnx` calls via OnnxRuntime
and various other features. We include `WordCountIntegrationTest` as an
example of how to call it on a standard wayang plan. We also provide
`setComplexityClass()` on `FunctionDescriptor`s that likewise will be used by
the new cost model.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
