Actions Panel
RAPIDS Academy - Easy Python Multi-GPU Programming with Dask-cuDF
Guest Matthew Rocklin (Dask, Coiled, ex-Nvidia) gets you going with Python multi-GPU programming in our latest live stream and optional lab.
When and where
Date and time
Tuesday, August 11, 2020 · 11am - 1pm PDT
Location
Online
About this event
UPDATE:
- Aug 18th @ 11a - 1p PT: New date for Dask To give Matthew and the Coiled team time to setup a special surprise for an even better session, they asked to delay by 1 week. Same time, same logistics, just the 18th instead of the 11th
- Aug 11th @ 11a - noon PT: Special interim session with BlazingSQL CTO Felipe Aramburu! We will still run a multi-GPU intro & hands-on (mixed BlazingSQL + Dask-cuDF), and hang out after on the RAPIDS Slack. If you missed the BlazingSQL session, or can't make next week's on Dask, or want to see all this together, Felipe is among the biggest contributors and most energizing figures in the RAPIDS ecosystem.
-------
INTERIM DASK+BLAZINGSQL SESSION
August 11th @ 11aPT - noon PT
Felipe Aramburu, the CTO of BlazingSQL, will discuss in-depth how out-of-core query execution works with BlazingSQL. We'll demonstrate single-node multi-GPU with Dask, before guiding everyone through a short set of queries that query datasets larger than available GPU memory.
See links below for the YouTube live stream channel, and for the tutorial, the Slack channel
----------
MAIN DASK SESSION
August 18th @ 11a PT - 1p PT
Welcome!
The third RAPIDS Academy live stream and optional lab shares how to get going with Python analysis on multiple GPUs:
- Learn about the RAPIDS ecosystem and its interconnected components
- Deep dive into single-GPU Python with cuDF data frames
- Start on dask-cudf, which uses the dask multiprocessing library to streamline writing multi-GPU Python and automatically break up datasets too big for GPU memory
- Tech: Includes Python Jupyter notebooks, cuDF, dask, and dask-cudf
GUEST INSTRUCTOR:
Matthew Rocklin
CEO, Coiled (Dask creator, ex-Nvidia)
FORMAT:
- Live Stream (40 min.): RAPIDS, Dask, and dask-cudf overview
- (20min break)
- Live Instructor-Led Lab (1hr): Optional with limited seats
- Free GPU, environment, and instructions will be provided, and open-source materials will be posted online.
INSTRUCTIONS:
Register for one of:
- The Live Webinar (40 min)
- The Live Webinar (40 min) + Instructor-Led Lab (1 hr)
Closer to the event, we will email + update the site with:
- The live video stream & group chat
- Instructor-led lab login & group chat instructions
- Optional instructions and materials for trying yourself
- Optional help chat for pointers if trying for yourself
FAQ:
Q: Is this free?
A: Yes!
Q: ... including the GPUs?
A: Yes! We will provide free GPU test lab environments and free DYI pointers
Q: Do I need to be a data science expert?
A: Nope! We expect basic scripting-level knowledge of Python, such as lists and loops. Experience with data systems like Python Pandas, SQL, or Spark will help.
Q: How will this work logistically?
A: Live stream:
Head over to the RAPIDS Academy channel on YouTube, which will open 5-10min ahead of time: https://www.youtube.com/channel/UCq3xirW77NvGvA1fNDMuJbQ
To chat, join the #RAPIDS_Academy channel on the RAPIDS Slack: https://rapids.ai/community.html
A: Optional lab:
We will email ticket holders a link to a private Zoom session and post materials to our Github. Day-of, we will run the talk and Q&A portion, take a 20min break, and switch to the private stream. Participants will be receive web logins to prepared data environments, and then instructors will work with them through the sequence of tasks.
Q: Will this be recorded?
A: The public parts! The live stream portion and notebooks will be published, but the lab will be kept private
Q: Will the answers be posted?
A: Yes, they will be made available at the same time as the labs
Q: How will you contact us, and how should we reach you?
A: We will email registered participants, and you can both email us back and use the form at LearnRAPIDS.com
Tags
About the organizer
RAPIDS Academy
Learn Python GPU coding and data science from the creators and top users of the Python RAPIDS.ai ecosystem. Our diverse online and in-person learning paths give analysts a friendly wayto learn what they need to know and, for data scientists, help them quickly ramp up with pydata GPU replacements. Pick general data tutorials, labs, and videos, or specialized to your field such as in security, marketing, and genomics.
Graphistry: 100X Investigations
The Graphistry GPU-accelerated visual investigation and automation platform helps startups, enterprises, and federal teams perform their data-intensive graph projects. We bring heavy experience with security & fraud investigation, and are growing in fintech, genomics, and sales/marketing analytics.