Ross Wightman

@wightmanr

Technology Doer and Dreamer

Vancouver, BC
Vrijeme pridruživanja: travanj 2012.

Tweetovi

Blokirali ste korisnika/cu @wightmanr

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

  1. prije 10 sati

    Added ImageNet validation results for 164 pretrained models on several datasets, incl ImageNet-A, ImageNetV2, and Imagenet-Sketch. No surprise, models with exposure to more data do quite well. Without extra, EfficientNets are holding their own.

    Poništi
  2. proslijedio/la je Tweet
    30. sij

    Are you an angel investor based in Vancouver interested in AI and AI-enabled companies? Join us at on February 5 for our themed meeting. Learn the latest trends from this hot industry, hear from a panel, and meet ~10 companies.

    Poništi
  3. 13. sij

    Another result, modifying the ‘deep stem’ from Bag of Tricks. Tiering the stem to (24, 48, 64) or (24, 32, 64) shows benefit over the D stem of (32, 32, 64). 77.6% D vs 78.0% for tiered on an SE-ResNeXt26-32x4d base. Tiered stem inspired by

    Prikaži ovu nit
    Poništi
  4. 13. sij

    I ran a few ResNet experiments over the holiday season. Testing my (re)implementation of Google’s AugMix, I trained a vanilla ResNet50 to 78.994% top-1, just ImageNet, 224 resolution. With test time mean-max pooling at 288x288, the same weights manage 80% top-1.

    Prikaži ovu nit
    Poništi
  5. 3. sij

    Kaggle + RL. Looking foward to seeing what is on the horizon for this new competition category!

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

    0/ Is there an Enterprise Margin Crisis? It's not uncommon to see software startups with surprisingly low margins (30-40%). We believe there is a broader trend going on here, which I explore in this thread.

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    13. pro 2019.

    Very excited to share the amazing libraries with the Vancouver data science community. Hope to see you there!

    Poništi
  8. 23. stu 2019.

    The TLDR of the paper; use adversarial examples as training data augmentation, maintain separate BatchNorm for normal vs adversarial examples. Neat. As usual I've ported & tested weights

    Prikaži ovu nit
    Poništi
  9. 23. stu 2019.

    Hard to keep up with and team! I just finished making it through EfficientDet and Noisy Student and now have AdvProp, w/ a quietly released update of all EfficientNet weights, incl a new B8 model spec.

    Prikaži ovu nit
    Poništi
  10. 21. stu 2019.

    More to throw on the ever increasing RL reading list...

    Poništi
  11. 15. stu 2019.

    A port of the recently (and long awaited) official Tensorflow release of MobileNetV3 large/small/minimalistic models added... I spent a long time staring at code before realizing the SE blocks don't inherit the act fn of their parent block as in MnasNet, EffNet, MixNet, etc.

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet

    To help accelerate 3D deep learning research, released Kaolin, a library that provides efficient implementations of differentiable 3D modules for use in DL systems.

    Poništi
  13. 30. lis 2019.

    Nice to see helpful AI / ML technology coming from local () inventors!

    Poništi
  14. 30. lis 2019.

    A implementation of EfficientNet-CondConv w/ some group conv crazyness and weights ported from official TF impl. Joining EfficientNet, EfficientNet-EdgeTPU, MixNet, and others for the only complete family of these models in PyTorch.

    Prikaži ovu nit
    Poništi
  15. 19. lis 2019.

    The last batch of weakly-supervised ResNext-101 models are great for transfer learning, I expect these will be too.

    Poništi
  16. 10. lis 2019.

    AMP (mixed precision) moving to core. Happy days! Thanks

    Poništi
  17. proslijedio/la je Tweet
    10. lis 2019.

    🎉 Introducing sotabench : a new service with the mission of benchmarking every open source ML model. We run GitHub repos on free GPU servers to capture their results: compare to papers, other models and see speed/accuracy trade-offs. Check it out:

    Poništi
  18. proslijedio/la je Tweet
    16. ruj 2019.

    2nd iteration of 🎉 We ditched our custom UI for an integration model w/ ur dev team's favourite tools e.g. 🕐 Real time running cost 🤖 Forecasting + anomaly detection w/ ML 🔧 Integrate your CI/CD and dev tools (e.g Jenkins) 🖌️ Fully customizable dash

    Poništi
  19. 25. srp 2019.

    MixNet in the mix now w/ ported weights working well. Thanks and team for another great model with weights and code!

    Poništi
  20. 24. srp 2019.

    Another impressive NAS result from the Google AI team.

    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:

    ·