Nicholas Malaya

@nicholasmalaya

Computational Scientist, AMD Research. Racing to Exascale is a team effort, but views on Twitter are my own.

Austin, TX
Vrijeme pridruživanja: rujan 2013.

Tweetovi

Blokirali ste korisnika/cu @nicholasmalaya

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @nicholasmalaya

  1. Prikvačeni tweet
    7. svi 2019.

    Thrilled to be a part of this historic effort.

    Poništi
  2. proslijedio/la je Tweet
    prije 15 sati

    AMD is proud to announce a new multi-year partnership with the Team! With our shared commitment to innovation and extreme performance, our partnership is a match made for the F1 track.

    Poništi
  3. proslijedio/la je Tweet
    3. velj

    I love Super Bowl statistics. (“No team that has failed to convert at least 35% of their 3rd downs has gone on to win” etc.) They are the best introduction to p-hacking that can be found outside major journals.

    Poništi
  4. 30. sij
    Poništi
  5. proslijedio/la je Tweet
    29. sij

    Wendell Misch: three stars nearby are known to be more ripe for supernova than Betelgeuse - Institute for Nuclear Theory – mjesto: UW: Physics & Astronomy Building

    Poništi
  6. proslijedio/la je Tweet
    21. sij

    . Annual Meeting attendees in Houston (1st week Feb), drop on by Thurs morning for the Frontier Tutorial! Learn about the architecture of 's Frontier supercomputer, GPU hardware, and the Frontier programming environment.

    Poništi
  7. proslijedio/la je Tweet
    15. sij

    Number of CPU cores in my machine 2012: 8 (Intel) 2019: 8 (Intel) 2020: 64 (AMD) 🔥

    Poništi
  8. proslijedio/la je Tweet
    13. sij

    Yamaha joins Tide in the "I can't believe you're making us say this, but Legal says we really have to say this" camp.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet

    I asked Tony Fadell about the iPod timeline for my fast project page. Summary: 😯.

    Poništi
  10. proslijedio/la je Tweet
    9. sij

    So proud that has announced the new Policy on Name Changes. To my knowledge, ACM is the first publisher with a policy like this, and it re-affirms that ACM is an inclusive community. This was a long term advocacy project, congrats to all!

    Poništi
  11. proslijedio/la je Tweet
    5. sij

    Consider: millions of years ago our antecedents gave a massive sacrifice of their left hemisphere. We lost a tremendous amount of short term memory and replaced it with Broca’s, Wernicke & the phonological loop. But why? So we can—talk. Thus chimpanzees can do this—we can’t:

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    5. sij

    In our tests, HIP-ifying our CUDA code has produced performant binaries for both AMD and NVIDIA archs. I was surprised and quite pleased by how easy the HIP interface was to learn and use.

    Poništi
  13. proslijedio/la je Tweet
    31. pro 2019.

    "An audacious Hollywood movie-style escape masterminded by his wife with the assistance of a Gregorian music band and a team of ex-special forces officers"

    Poništi
  14. proslijedio/la je Tweet
    29. pro 2019.
    Odgovor korisniku/ci

    The sad thing is that the ideal batch size is 1, and the ideal network architectures are different from the most popular ones. The reason we use large batches and ResNet architectures is the design of current Nvidia GPUs that force us to reduce everything to matrix products.

    Poništi
  15. proslijedio/la je Tweet
    27. pro 2019.

    In 1935, after the famous Einstein-Podolsky-Rosen paper arguing that quantum mechanics was incomplete, Schrödinger wrote to Einstein: "I am very happy that in the paper just published in Physical Review you have evidently caught dogmatic q.m. by the coat-tails". (1/n)

    Prikaži ovu nit
    Poništi
  16. proslijedio/la je Tweet
    26. pro 2019.

    Bayesian methods are *especially* compelling for deep neural networks. The key distinguishing property of a Bayesian approach is marginalization instead of optimization, not the prior, or Bayes rule. This difference will be greatest for underspecified models like DNNs. 1/18

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    15. pro 2019.

    NATIONAL CHAMPIONS!!!! HOYA SAXA!!! WAY TO GO, !!!!

    Poništi
  18. proslijedio/la je Tweet
    12. pro 2019.

    RAJA now has a support for AMD GPUs via a new HIP backend! Currently in our `develop` branch, and will be included in the next release

    Poništi
  19. proslijedio/la je Tweet
    8. pro 2019.

    The way Swan makes his statements clearly demonstrates that we will not see Intel 7nm before Q4 2022... 🙄 The worst is yet to come...

    Poništi
  20. proslijedio/la je Tweet
    28. stu 2019.

    Academia, get your shit together. It is unacceptable for grants and funding sources not to recognise parental leave or any other justified career break. Just the tip of the iceberg but seems like and easy one start fixing.

    Poništi
  21. proslijedio/la je Tweet
    21. stu 2019.

    Congrats to the 2019 ACM Gordon Bell Prize winning team at led by ! The team modeled a 10,000-atom 2D slice of a transistor on Summit, with a sustained performance of 85.45 PF for double precision and 90.89 PF for mixed precision. ➡️

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·