chronon
chronon copied to clipboard
airbnb
Chronon is a data platform for serving for AI/ML applications.
## Summary Adds an APPROX_HISTOGRAM_K operation based on the FrequentItems Sketch: https://datasketches.apache.org/docs/Frequency/FrequentItemsOverview.html This is mostly a wrapper on top of the frequent items sketch, but have made the two following...
## Summary Adds operations for [skew](https://en.wikipedia.org/wiki/Skew) and [kurtosis](https://en.wikipedia.org/wiki/Kurtosis). The implementation uses the same Welford/Knuth approach as the variance operation. https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Higher-order_statistics I have the operations set as non-deletable as the reversals...
## Summary In this PR, we will check if there is a mutation table in the entity source to determine if it is a mutation source. The entity source itself...
## Summary We are supporting metadata upload to k-v store for key-value pair key->conf right now. We want to add a general class metadata endpoint to support more potential use...
## Summary Documentation updates for approx_histogram_k, skew, and kurtosis aggregations: https://github.com/airbnb/chronon/pull/735 https://github.com/airbnb/chronon/pull/740 ## Why / Goal ## Test Plan - [ ] Added Unit Tests - [ ] Covered by...
# CHIP-9: Support Model-based Transformations in Join & Chaining ## Problem Statement Model Inference is an important primitive form of transform function that ML practitioners use in creating feature pipelines....
## Context Our API to create objects currently takes all arguments at once. We use this style to create objects such as Source, GroupBy, Join etc. *We are essentially exposing...
Large changes or additions to Chronon require a detailed proposal. Below is a template for the proposals. CHIP has a more obvious pronunciation than CIP. This template is a recommendation...
# Problem Depending on the parallelism, we are simultaneously running several joinPart jobs. Besides that, we also have bootstrap and the final join potentially running in parallel. In the future...