Andrew Kostousov

Improve documentation

1

We need to describe all Vektonn features with detailed usage examples for major ones. Also it would be nice to have documentation page with service architecture overview. Proposed navigation structure:...

The following code: ``` var serializer = new Serializer(new AllPropertiesExtractor()); var dict = new Dictionary { {("k11", "k12"), "v1"}, {("k21", "k22"), "v2"} }; var bytes = serializer.Serialize(dict); serializer.Deserialize(bytes); ``` fails...

1. Test [TestSwitch.TestString()](https://github.com/skbkontur/GrobExp.Compiler/blob/a690f379afbdc28614f8235cf2d9c635cf7f0b8c/GrobExp.Compiler.Tests/TestSwitch.cs#L149) is flaky on dotnet core. 2. Test [TestPerformance.TestSwitch2()](https://github.com/skbkontur/GrobExp.Compiler/blob/a690f379afbdc28614f8235cf2d9c635cf7f0b8c/GrobExp.Compiler.Tests/TestPerformance.cs#L602) fails constantly on dotnet core.

We have some perf [tests](https://github.com/skbkontur/GrobExp.Compiler/blob/a690f379afbdc28614f8235cf2d9c635cf7f0b8c/GrobExp.Compiler.Tests/TestPerformance.cs#L20) but they are far from ideal. Basically the goal is to measure perf difference between GrobExp.LambdaCompiler.Compile() and native Expression.Compile() when running in different environments (OS,...

Optimize DTOs json serialization

2

[test_perf_serialize](https://github.com/vektonn/vektonn-client-python/blob/10d6311fe00a496bb6c575fa1854a91f8f780950/tests/test_dtos_perf.py#L29) run shows that [`VektonnBaseModel.json()`](https://github.com/vektonn/vektonn-client-python/blob/10d6311fe00a496bb6c575fa1854a91f8f780950/src/vektonn/dtos.py#L17) is a huge bottleneck: ``` ----------------------------- Captured stdout call ----------------------------- 28285893 function calls in 25.275 seconds Ordered by: internal time ncalls tottime percall cumtime percall...

[ScaNN] Are there any plans to release tf-serving-scann Docker image corresponding to ScaNN 1.2.10 release?

1

The last official ScaNN TF Serving Docker image is [v2.11.0](https://hub.docker.com/layers/google/tf-serving-scann/2.11.0/images/sha256-dbbaeb571883546ca6ea9f7eeba58917a030ae8f7ff0bf1a814fa68b3ded4e55?context=explore) and it dates back to Dec 2022. Is it compatible with models built by ScaNN [v1.2.10](https://pypi.org/project/scann/1.2.10/), which was released recently?