Skip to main content

Weekly Meeting Minutes 2021-01-05

· One min read
Mahout TeamMahout Team
Apache Mahout

Andrew showed Trevor how to file his first board report.

Springer Publishing has shown interest in a second edition of the Encyclopedia of Big Data, and Musselman plans to update the chapter on Mahout.

Trevor needs to make a git repo to share with his experiments of Python+Mahout.

Reminder- community calls are open to all. We'll try to get a good link and post it to one of the minutes. In the mean time join dev@mahout.apache.org for the weekly link.

Weekly Meeting Minutes 2020-12-15

· One min read
Mahout TeamMahout Team
Apache Mahout

Trevor able to do Mahout in-core matrix addition in Python. Will continue working, but good first step. Want to see POC that DRMs will also "work" and then will sketch plan.

Weekly Meeting Minutes 2020-12-08

· 2 min read
Mahout TeamMahout Team
Apache Mahout

New Meeting time is Tuesdays, still hammering down final time.

Musselman to continue work on 2130 (adding talks page).

Trevor making (slow) progress wtih Py4J / Numpy / Python bindings (life getting in the way, but seems doable).

Palumbo interested in FPGA-

Investigation: FPGA4Mahout, Mahout BLAS Subroutines on FPGA. we are considering integration of the fBLAS library for FPGA acceleration of BLAS subroutines on mahout. fBLAS accepts JSON expressions of algebraic subroutines and generates OpenCL code with Just In Time compilation and FPGA flashing. We are first considering a Naive approach upon evaluation to convert the computation graph of a mahout expression into JSON, compile and Flash the FPGA and stream data through the FPGA in SIMD vectored fashion.

The fBLAS library shows significant speedups over CPU BLAS operations. If we find significant performance gains, In conjunction with the current and ongoing effort to implement Python Bindings for Mahout, this could attract HPC developers with needs for near-real-time computation, by providing Python and Java bindings for FPGA accelerated matrix algebra available for distributed and in-core math.

An end goal of this effort would be to provide scientists, engineers, and others with little to no Hardware experience a Zeppelin Notebook on which they could can develop models or circuits, using the mahout DSL, while concurrntly programming the model onto an FPGA"

Weekly Meeting Minutes 2020-11-13

· One min read
Mahout TeamMahout Team
Apache Mahout

Weekly community meeting minutes.

Collected Videos / Conference Talks

Andrew Musselman is going to try to work on MAHOUT-2130 (Build talks page) this weekend. If you have / know of any content, please send it to him or just add it after he launches the page.

Issue Grooming

The following issues are all related

  • MAHOUT-2122 Create getting started on Docker page
  • MAHOUT-2134 Update Mahout in Zeppelin page
  • MAHOUT-2129 Rebuild website docs, add Getting started with Zeppelin to Nav
  • MAHOUT-2126 Update Zeppelin Containers
  • MAHOUT-2133 Update "Getting Started in Mahout" page.

The narrative is- we should update (or replace) the getting started in Mahout page with Getting Started with Mahout in Zeppelin. The Zeppelin Docker should have some more "Getting Started" Notebooks. The current Mahout in Zeppelin page should point to the new Mahout in Zeppelin on Docker page. The nav should be updated to reflect all of this.

Weekly Meeting Minutes 2020-11-06

· One min read
Mahout TeamMahout Team
Apache Mahout

Weekly community meeting minutes.

Topic: Apache Mahout Wikipedia Page

Andrew Palumbo added to the Native solvers sections- thanks Andy!

Joe Olsen going to take some cracks at it next. (all are welcome to!)

Topic: Starting on Python Bindings

No progress to date- weather is unseasonably nice in Chicago.

Topic: K8s Spark+Docker container

  • Joe going to create some docs on Mahout + K8s based on Trevor's Kubeflow book examples.
  • Trevor's going to create Kubeflow + Mahout example based on book. (Someday)

Created

  • MAHOUT-2133 Update Quickstart Page
  • MAHOUT-2134 Update Zeppelin Tutorial

Weekly Meeting Minutes 2020-10-30

· One min read
Mahout TeamMahout Team
Apache Mahout

Topic: Apache Mahout Wikipedia Page

Trevor has been doing some serious overhauling on the Apache Mahout entry on Wikipedia.

However, he'd appreciate it if people could fill out sections, especially if they have expertise.

Further- we would like to add architecture diagrams- both the original and the one with the native solver- need to ask what the copyright it on those/if we can use them.

Topic: Zeppelin Docker Image

The Apache Zeppelin+Mahout docker image is live- you can get started with Mahout now by following the instructions here

Topic: Starting on Python Bindings

Trevor is going to start working on Python bindings. He thinks adding Mahout as a library for pyspark will significantly increase adoption. If anyone has experience using py4j or is otherwise interested in helping or wants to give words of encouragement reach out on dev or email him directly.

14.1 Is Released!

· One min read
Mahout TeamMahout Team
Apache Mahout

Oh happy day! A lot of work went into this release with getting the build system to work again so that we can release binaries. There really weren't too many new features over 14.0, but binaries now exist in Maven.

On another note, we had a really exciting "MahoutCon" track at ApacheCon@home this year. A special thanks to all of our speakers, and we plan to post the videos here when we're able.

Code Freeze and Wierd Numbering

· One min read
Mahout TeamMahout Team
Apache Mahout

On wierd version numbering...

We were under pressure to release a "version above 1.0" which seemed sort of arbitrary to us.

We decided to just stick with our original schema, but take it x10. Our last release was 0.14.0, and thus this release is 14.1. So if you're wondering where versions 1.0-13.0 are, that's what happened.

14.1 is a long overdue fix for us releasing binaries for to use, instead of having to build your own. 0.14.0 was us finally moving the old Map Reduce stuff out of the main trunk, because we haven't been about that for quite some time.

If you would like to help us test the release candidate, send an email to dev-subscribe@mahout.apache.org and reply to the confimration email to join our mailing list, and happy Mahouting!

Preparing for version 14.1

· One min read
Mahout TeamMahout Team
Apache Mahout

Hey all! We're preparing for Mahout version 14.1.

This release might not seem super exciting from a user perspective (except we're bumping Apache Spark to version 2.x and Scala to version 2.11) but it includes some very important 'behind the scenes changes'. We're doing a heavy refactoring which will combine math and math-scala into a module called core (no change to code required!).

We will be moving H2o and Flink-batch engines into community that is to say the code will still be available, but we won't be actively maintaining it (we will accept PRs though). We're moving the Map Reduce code to community and have considered to once again accepting PRs against this code (if you have an open PR, we'll hit you up after the release about rebasing and considering to resubmit the PR)

We're going to be updating the Java dependency from 1.7 to 1.8.

There was a LOT of old junk in the POMs which we have clipped out.

Stay tuned and feel free to help out!