| commit | 8dbfc235cc222b5e7b3fdf9adfc1a09b95c84249 | [log] [tgz] |
|---|---|---|
| author | baunsgaard <baunsgaard@tu-berlin.de> | Mon Aug 07 13:37:40 2023 +0200 |
| committer | baunsgaard <baunsgaard@tu-berlin.de> | Tue Aug 08 14:59:33 2023 +0200 |
| tree | 828282cf4469d9f951340038150ac3ff33c8f8b2 | |
| parent | 68c2c17db56c8c9ea5b1985bbf3b525d4ac8d022 [diff] |
[SYSTEMDS-3608] Cocode shortcut This commit adds a shortcut in que based cocode algorithm that indicate if no cocoding have happened the last 5 tries in the que abort cocoding. Previously cases where cocoding was on the border then all pairs of columns would be tried single threaded. To avoid this the implementation is now avoiding the all pairing if no cocoding is happening. The cost optimization of cocoding have been tuned to add a minor delta on the cost based on the average of the column indexes in the groups. This makes the que based optimizer likely to cocode columns next to each other if they have similar cost, making the cocoding algorithm faster in cases with high corelation between columns next to each other. This commit also adds an TwoRangesIndex, that is useful in the case where almost all columns get cocoded, and the greedy algorithm combine colgroups with thousands of groups. In this case the column arrays combining dominated. Also added in this commit is compression schemes for SDC and refinements for SDC, and a natural update and upgrade scheme progression from initially empty columns to const to DDC. Next up is to add a transition to SDC in case the distribution of values become dominated by specific values. We also add a minor update to AggTernaryOp for compressed with a shortcut that avoids an decompression in L2SVM. Closes #1874
Overview: SystemDS is an open source ML system for the end-to-end data science lifecycle from data integration, cleaning, and feature engineering, over efficient, local and distributed ML model training, to deployment and serving. To this end, we aim to provide a stack of declarative languages with R-like syntax for (1) the different tasks of the data-science lifecycle, and (2) users with different expertise. These high-level scripts are compiled into hybrid execution plans of local, in-memory CPU and GPU operations, as well as distributed operations on Apache Spark. In contrast to existing systems - that either provide homogeneous tensors or 2D Datasets - and in order to serve the entire data science lifecycle, the underlying data model are DataTensors, i.e., tensors (multi-dimensional arrays) whose first dimension may have a heterogeneous and nested schema.
Quick Start Install, Quick Start and Hello World
Documentation: SystemDS Documentation
Python Documentation Python SystemDS Documentation
Issue Tracker Jira Dashboard
Status and Build: SystemDS is renamed from SystemML which is an Apache Top Level Project. To build from source visit SystemDS Install from source
