RISE Camp 2020

RISE Camp was held remotely on October 29-30, 2020.

RISE Camp ​is ​a bootcamp ​organized ​by ​the UC ​Berkeley ​RISELab ​where ​you ​can get exposure to research and ​hands-on ​experience ​with systems and technologies for ​emerging AI applications including reinforcement learning, prediction serving, agile ML development, context management, and AI security. ​

Questions? Please email: risecamp@cs.berkeley.edu

Videos

Agenda

DAY 1 – THURSDAY, OCTOBER 29th 2020 (All times in Pacific Daylight Time)

  • 9:00 – 10:15 AM:
    • Ray (Stephanie Wang) : Attendees will learn how to use Ray to build a scalable web app for movie recommendations based on online learning. Slides available here.
  • 10:15 – 10:30 AM: Break
  • 10:30 – 11:45 AM:
    • Modin (Devin Petersohn): Attendees will learn how to use Modin in different environments, and the key components of the system. Slides available here.
  • 11:45 – 12:45 PM: Lunch with Office Hours (Prof. Ion Stoica)
  • 12:45 – 2:00 PM:
    • Autopandas (Rohan Bavishi): Attendees will learn how a rich specification that goes beyond input-output examples can be both convenient and easier to provide, and still benefit synthesis of table transformations. The tutorial will include a demonstration of our special UI to support this specification and some example tasks we can solve using the UI. Slides available here.
  • 2:00 – 2:15 PM: Break
  • 2:15 – 3:30 PM:
    • NBsafety (Stephen Macke): Attendees will learn how NBSafety can make it easier to reason about out-of-order execution in notebooks (thereby avoiding a common class of bugs that commonly occur in notebook environments) by highlighting cells with fresh / stale variables. Slides available here.

DAY 2 – FRIDAY, OCTOBER 30th 2020 (All times in Pacific Daylight Time)

  • 9:00 – 10:15 AM:
    • MC2 (Chester Leung): Attendees will learn how to use MC2 for secure collaborative learning. That is, mutually distrustful data owners can use MC2 to jointly train a model on their data, but without revealing their individual data to each other. Slides available here.
  • 10:15 – 10:30 AM: Break
  • 10:30 – 11:45 AM:
    • ERDOS (Sukrit Kalra): Attendees will learn how to develop a planning component for an autonomous vehicle and drive it in simulation. Further, the tutorial will also show how the ability of ERDOS to dynamically adapt to the environment allows an autonomous vehicle to prevent accidents. Slides available here.
  • 11:45 – 12:45 PM: Lunch with Office Hours (MC2)
  • 12:45 – 2:00 PM:
    • AdaHessian and PyHessian (Amir Gholami and Zhewei Yao):Attendees will learn advanced analysis of neural networks by learning to compute second-order derivatives. The tutorial would then show how this could be used to (i) analyze the loss landscape of the model,  and (ii) how it could be used to achieve high accuracy training of different neural networks. Slides available here.
  • 2:00 – 2:15 PM: Break
  • 2:15 – 3:30 PM:
    • LUX  (Doris Lee): Attendees will learn how Lux supports fast and easy experimentation with data through visualizations seamlessly in their Jupyter notebooks. Slides available here.

FAQ

What are the prerequisites for attendees of the RISE Camp?

  • “Modern” browser (Firefox, Chrome, Safari etc.)
  • Zoom account. PLEASE NOTE: All participants and hosts are now required to sign into a Zoom account prior to joining meetings hosted by UC Berkeley. Participants who do not have a UC Berkeley-provided Zoom account can use a Zoom account provided by their institution or create a free, consumer Zoom account (at https://zoom.us/freesignup/). If your company policies do not allow downloading Zoom on your laptop, please choose “Join from your browser” option when prompted. For best results, we recommend using Chrome with Zoom add-on
  • Experience programming in Python in notebook environment
  • Basic understanding of AI/ML concepts (e.g., training, validation, linear models)

Tutorials

Ray:

You can find the code for the tutorials at this GitHub repo.

Also, here are some instructions for how to run the tutorials locally using Docker. You can also run the code directly from the GitHub repo included above, but the Docker image already has the Python and data dependencies pre-packaged, so we recommend running with Docker.

  1. Download the Docker image from Docker Hub (make sure you have Docker installed first):
    $ docker pull swangster/rise-camp-tutorial:v0
  2. Start a container. This command includes a flag to forward the default port for Jupyter.

$ docker run -it -p 8888:8888 swangster/rise-camp-tutorial:v0

  1. Inside the container, navigate to the directory with the Jupyter notebooks and start Jupyter lab:

$ cd /root/rise-camp-tutorial && jupyter lab –ip=0.0.0.0 –allow-root

  1. The last command should have printed a link that you can click on or copy-paste into a web browser to navigate to the Jupyter Lab session.

Modin:

To run the tutorial, do the following:

git clone https://github.com/modin-project/modin.git

cd modin

pip install virtualenv

virtualenv modin-tutorial

source ./modin-tutorial/bin/activate

cd examples/tutorial/

pip install -r requirements.txt

jupyter lab

Autopandas:

Link to repo : https://github.com/rbavishi/gauss-rise-camp

NBSafety:

The NBSafety tutorial + instructions are available at this GitHub repo.

MC2 :

You can find the tutorial with instructions here: https://github.com/mc2-project/risecamp.

ERDOS

To run the tutorial on your own machine, we require Docker and recommend a good GPU in order to run the simulations. Follow the following instructions to run the image:

docker pull erdosproject/pylot:risecamp

Nvidia-docker run -itd –name pylot_risecamp_20 -p 80:8080 erdosproject/pylot:risecamp /bin/bash

And then navigate to localhost to see the jupyter notebook!

AdaHessian and PyHessian:

You can find the tutorial with instructions here: https://github.com/yaozhewei/PyHessian_AdaHessian_RiseCamp

Lux:

You can find the tutorial instructions here:  https://github.com/lux-org/lux-binder/

The Github repo contains instructions on how to set up the tutorial locally using Docker.

You can also try out the tutorial by accessing a live notebook at this Binder link: https://mybinder.org/v2/gh/lux-org/lux-binder/master?urlpath=tree/exercise