Live data from Hacker News

Viewing profile — mattjjatgoogle

mattjjatgoogle

HN member
Joined
Sun, Feb 24, 2019, 3:30 PM UTC
HN karma
77
Public activity
14 items

About mattjjatgoogle

No profile information was provided.

Recent public activity

  1. comment
    Comment #42938115

    An author's tweet thread: https://x.com/jacobaustin132/status/1886844716446007300

  2. story
  3. comment
    Comment #37741569

    Actually that never changed. The README has always had an example of differentiating through native Python control flow: https://github.com/google/jax/commit/948a8db0adf233f333f3e5…

  4. comment
    Comment #37741506

    Actually, that's never been a constraint for JAX autodiff. JAX grew out of the original Autograd ( https://github.com/hips/autograd ), so differentiating through Python control flo…

  5. comment
    Comment #37706167

    You're right! Maybe we should revise that... I made https://github.com/google/jax/pull/17851 , comments welcome!

  6. comment
    Comment #37706060

    Have you seen JAX MD? https://github.com/jax-md/jax-md

  7. comment
    Comment #34931579

    You're right that downstream libraries have often tended to introduce magic (some more than others), and moreover one library's magic is typically incompatible with other libraries…

  8. comment
    Comment #34930556

    Thanks for taking the time to explain these. > It's been a bit, but I think the most frustrating errors were around mapping pytrees (like this issue https://github.com/google/jax/i…

  9. comment
    Comment #34929340

    If you have any particular examples in mind, and time to share them on https://github.com/google/jax/issues , we'd love to try to improve them. Improving error messages is a priori…

  10. comment
    Comment #34929121

    Thanks for the kind words! We've been doing a lot more work in this direction too, for both compiler-based automatic parallelization [0] and a work-in-progress pmap successor for '…

  11. comment
    Comment #34757552

    I'm sure there's a lot of good material around, but here are some links that are conceptually very close to the linked Autodidax. (Disclaimer: I wrote Autodidax and some of these o…

  12. comment
    Comment #24549195

    The difference is that in TF1 you had to use tf.cond, tf.while_loop etc for differentiable control flow. In JAX you can differentiate Python control flow directly, e.g.: In [1]: fr…

  13. comment
    Comment #22596615

    No worries! I didn't mean it as a correction so much as just a discussion; I'm sure it's true that other autodiff systems have very sophisticated automatic remat (like https://open…

  14. comment
    Comment #22596180

    Rematerialization in autodiff is super interesting! XLA does rematerialization optimizations, so you get those automatically under jax.jit. There's also the jax.checkpoint decorato…