commit | ebcc44db28d59ec076e3a56cee1051361a1d3c24 | [log] [tgz] |
---|---|---|
author | Matthias Boehm <mboehm7@gmail.com> | Sun Jan 31 19:40:36 2021 +0100 |
committer | Matthias Boehm <mboehm7@gmail.com> | Sun Jan 31 20:21:42 2021 +0100 |
tree | 8d8662ccc502ae23c2b20e56709d538a2e4dac7c | |
parent | 6396846748f2e80ade7c04aae498e83b811cac5a [diff] |
[SYSTEMDS-2819,2020] Various ctable improvements (rewrites, spark ops) * New ctable-reshape rewrite to avoid unnecessary intermediates (in CP, this also enables large datasets w/ nrow*ncol > max-integer) * Improved estimation of ultra-sparse distributed matrices to avoid huge number of partitions on ctable and other operations (on criteo day 21, 11K vs 500K partitions) * New ctable parameter to specify need to emit empty output blocks (on ultra-sparse matrices these empty blocks dominate the total size and are only needed for sparse unsafe distributed operations, right now this is an undocumented parameter, in the future this should become an interesting property and be propagated across the entire program) * Better error handling in spark ctable instructions to indicate invalid output dimensions (e.g., invalid pre-pass finds 0 max dimension value due to to missing values) * Avoid unnecessary partitioning on parfor entry (despite expected zipmm/cpmm) if distributed matrices are already hash-partitioned. * Leverage new ctable configurations in slice finder built-in * Fix DMLScript error printing to avoid NPEs (on-existing default opts)
Overview: SystemDS is a versatile 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