Tweetovi

Blokirali ste korisnika/cu @slurpwaffl

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

  1. Prikvačeni tweet
    31. sij 2016.

    more stuff posted to bandcamp free downloads as always

    Poništi
  2. 3. velj

    RuntimeError: size mismatch

    Poništi
  3. 2. velj
    Prikaži ovu nit
    Poništi
  4. 2. velj

    wouldn't it be great if instead of having desktops, we had a dock with gigabit + 3-4 usbc for monitors, and you put your phone on the dock and connect to your personal cloud desktop. then any intensive work you do could easily be scaled to new nodes

    Poništi
  5. 1. velj

    thanks u and and probably a bunch of other people. btw have there been any big performance benchmarks of tracker vs zygote/diffeqflux?

    Prikaži ovu nit
    Poništi
  6. 1. velj

    it's really exciting to see Flux.jl make the switch to Zygote.jl. this is a big step for autodiff. the new NeuralODE interface in DiffEqFlux.jl is also fantastic.

    Prikaži ovu nit
    Poništi
  7. 30. sij

    more quotes john v: "There are some phonemes that you can’t get to directly from a given starting phoneme" "So you quantify metabolic cost and then make your loss function some composite of that and information transfer"

    Prikaži ovu nit
    Poništi
  8. 30. sij

    quotes from john v: "people tend to say that your f0 is determined by the anatomy of your glottis insofar as it takes more energy to move a heavy one", "if you knew the mass of the vocal folds, then it wouldn’t be hard to measure the amount of energy it takes to move them"

    Prikaži ovu nit
    Poništi
  9. 30. sij

    has anyone built any nlp models that are restricted to the same conditions that humans have in creating sound? i think it would be interesting to see what patterns emerge in a task restricted by the amount of energy it takes to make certain sounds

    Prikaži ovu nit
    Poništi
  10. 25. sij

    Go and conways game of life are essential components of any free-energy sustained entity

    Poništi
  11. proslijedio/la je Tweet
    23. sij

    hello qubits

    Poništi
  12. 19. sij

    trained a vae to compress some of my songs, its pretty neat to hear it learn sound: code:

    Prikaži ovu nit
    Poništi
  13. 5. sij

    using a laser to create micro-bots is a phenomenal idea. AI will be the catalyst to enable assembly of truly incredible devices. huge potential uses on mars

    Poništi
  14. 5. sij

    google translate on desktop needs to have a conversational feature, like there is on mobile. needs to add features that easily enable using translate as a language learning tool.

    Poništi
  15. 4. sij

    great work to all the people who participated in from . i'm so excited to see this work start to improve RF allocation!

    Poništi
  16. 4. sij

    or rather, ask if you'd like to add it to your google calendar

    Prikaži ovu nit
    Poništi
  17. 4. sij

    gcal needs better chrome integration. chrome searches should work more like google assistant. if you want to add event(s) to calendar, just use chrome search, if there are descriptions, locations etc, it should add it

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet

    In NH reacts to refusing to commission early state polls: "The DNC saying we can do nothing about it seems disingenuous bc they can clearly just reach out to any of the approved polling organizations & ask for a poll to be conducted in a given timeline."

    Poništi
  19. 1. sij
    Poništi
  20. 28. pro 2019.

    all these sites that use ml/ai heavily should have a slider for learning rate. i feel the algos overfitting on pretty much everything

    Poništi
  21. 27. pro 2019.

    arch linux install bad

    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:

    ·