[TWIM Notes] Sep 28 2020

Message boards : News : [TWIM Notes] Sep 28 2020
Message board moderation

To post messages, you must log in.

AuthorMessage
pianoman [[email protected] Admin]
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Jun 20
Posts: 356
Credit: 7,876,185
RAC: 77,595
Message 537 - Posted: 29 Sep 2020, 0:47:41 UTC

This Week in [email protected]
Notes for Sep 28 2020
A weekly summary of news and notes for [email protected]

Summary
Wow, what a week. It's been 7 days since we released Dataset 3 WUs, and YOU, our volunteers have already crunched 48 of the workunits towards our first milestone (100 networks for each of the 100 automata). Client v9.61 appears to be working well, Dataset 1+2 WUs continue working in the background. Dataset 3 WUs were coming in so fast, we had optimize the validation process because server was getting overwhelmed!

At this rate, we're going to be ready for the first release of Dataset 3 (milestone 1) before Daatset 1+2 complete. That's a good result. We'll also release more Dataset 3 WUs soon, This also bodes well for Dataset 4.

Preparations for Dataset 4 (MNIST and TrojAI) are underway. Paper writing still underway, it is unlikely we'll make the deadline we were targeting for the end of this month, because we feel including Dataset 3 would be an even more compelling result.

In other good news, the new server has shipped from the vendor, and should be arriving at the university tomorrow! Meaning sometime in the next few weeks there will be some downtime while we transition to a new server with twice the ram, twice the speed, 3 times the number of cores, and 8 times the storage.

News:

  • Dataset 3 WUs processing going fantastically, uch faster than anticipated.
  • We'll do an official release of a preliminary Dataset (1+2) once we have at least 1000 examples of each machine type, and we're getting closer! Those WUs continue working in parallel with Dataset 3.
  • New server should arrive this week.
  • Badges should be rolling out this week too.
  • Training/eval data for Dataset 4 has been created, and work continues on updating the client and validation process to handle the image-based dataset



Project status snapshot:

Tasks
Tasks ready to send 2851
Tasks in progress 35721
Users
With credit 807
Registered in past 24 hours 30
Hosts
With recent credit 1994
Registered in past 24 hours 20
Current GigaFLOPS 30112.54

Dataset 1 and 2 progress:

SingleDirectMachine      10002/10004
EightBitMachine          10001/10006
SingleInvertMachine      10001/10003
SimpleXORMachine         10000/10002
ParityMachine              723/10005

ParityModified             170/10005
EightBitModified          5747/10006
SimpleXORModified        10005/10005
SingleDirectModified     10004/10004
SingleInvertModified     10002/10002 

Dataset 3 progress:
Milestone 1, 100x100:  4854/10112


Last week's TWIM Notes: Sep 21 2020

Thanks again to all our volunteers!

-- The [email protected] Admins
ID: 537 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
bozz4science

Send message
Joined: 9 Jul 20
Posts: 105
Credit: 2,157,126
RAC: 18,827
Message 539 - Posted: 29 Sep 2020, 14:37:34 UTC - in response to Message 537.  

Just checking back in here at [email protected] Awesome read. Honestly, I am also very stunned with the speed the crowd is generating here. Has there been a recent bump in # of hosts / members that accounts for this speedup?

Just got my hand on a few 3rd run rand_ WUs, that do take considerably longer on my system than the prior 1/2 WUs. Even though, I am excited for the intermediate results and the upcoming launch of our 4th experiment.

Thrilled to read about the server upgrade and the backend optimizations. In short, well done!

Good luck on the paper!

ID: 539 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : News : [TWIM Notes] Sep 28 2020

©2021 [email protected] Team
A project of the Cognition, Robotics, and Learning (CORAL) Lab at the University of Maryland, Baltimore County (UMBC)