Training large language models on HPC-Systems ⚙ At the ISC High Performance Conference 2023 in Hamburg last week, a OpenGPT-X project poster was presented by the Accelerating Devices Lab (X-Dev) of OpenGPT-X partner Forschungszentrum Jülich ➡ https://t.co/WiZI2HOcZO
I NEED EVERY SINGLE ONE OF MY FOLLOWERS TO RT THIS
70,000 UCU members strike tomorrow.
Every university in the UK.
Tomorrow we make history.
#ucuRISING
Fraction of permanent positions at research Universities
Switzerland 26%
Germany 27%
Netherlands 43%
Austria 45%
USA 37% + 15% TT
England 66%
France 67%
1/5
I had a great time at the 14th JLESC workshop hosted by National Center for Supercomputing Applications
There were interesting talks and connections, for example in our Women in High Performance Computing (WHPC) sessions.
I used…https://t.co/sGna4cack7 https://t.co/50K0lSow2W
Since 1969 Strassen’s algorithm has famously stood as the fastest way to multiply 2 matrices - but with #AlphaTensor we’ve found a new algorithm that’s faster, with potential to improve efficiency by 10-20% across trillions of calculations per day! https://t.co/nLvFbEDBuO
Let's exchange ideas and best practices about how local networks such as #JuWinHPC@fzj_jsc, #WiCS@BSC_CNS and Women@NCSA @NCSAatIllinois can support @women_in_hpc. Looking forward to hosting a breakout session the JLESC workshop this Thursday. Stream: https://t.co/udGbsm2scN
#JuWinHPC will be on the agenda of the JLESC workshop on 29.09. and streamed live:
▶️Talk by @SBroudeGeva (09:00 CT / 16:00 CEST)
▶️Discussion with @NCSAatIllinois and @BSC_CNS on how to establish such networks (10:00 CT / 17:00 CEST)
🗓️More info: https://t.co/yIvBNvofdr
Dear PhD students in Germany. A new ruling by the Federal Labor Court says: Recording time is mandatory.
The thread by @PschorrSimon indicates that this means, you have to be paid for over hours. How does that work in practice? Let's wait and see for a few more days...
Massive status bias in peer review.
534 reviewers randomized to review the same paper revealing the low status, high status, or neither author. 65% reject low status, 23% reject high status.
Amazing work by Juergen Huber and colleagues. #prc9
Hot take: When you have too many tabs open, at some point you should just impulsively close everything.
If you love something let it go. If it comes back it was meant to be; if it doesn't it never was.
@AndiH But what do you do instead to get that feeling of breaking chains holding you back, like your past does not determine your future, when actually you still have a presentation to prepare for the next day?