<aside> πŸ“’ [2022.09.08] All clusters are deployed!! Thank you for waiting us.

</aside>

Why Should We Use a Resource Manager?πŸš€

Equivalent question: β€œWhy should we learn boilerplates about requesting GPUs while all we want to do is just running a single .py file?”

Of course, it is much easier to use when we assign each GPU to each user for a whole semester. But we should keep in mind that

  1. There are far more users than GPUs.
  2. Graduate students typically make use of multi-GPU codes.

Besides, last semester, when we didn't do any time-multiplexed management, averaged GPU utilization was under 1%!! Therefore, it's natural to reach the consensus that we should do some fine management to use limited resources as efficiently as possible. β€œThe greatest happiness of the greatest number.”

Tutorials

PPTs have some crucial information. Please login with KHU-MS account if you want to see. The account will end with β€˜@office.khu.ac.kr’

0. Introduction

1. Practice

Get Access

Sign Up First

KHU Seraph

Access with terminal

# inside the KHU network
ssh YOURNAME@CLUSTERDOMAIN
# ex) ssh [email protected]

# outside of the network (for foreigners and commuters)
ssh YOURNAME@CLUSTERDOMAIN -p 30080
# ex) ssh [email protected] -p 30080

[EXTREMELY HIGHLY RECOMMENDED] Join our Slack channel