Quintin Pope comments on QAPR 5: grokking is maybe not that big a deal?

Quintin Pope 25 Sep 2023 19:46 UTC
2 points
0
It can be induced on MNIST by deliberately choosing worse initializations for the model, as Omnigrok demonstrated.
- gpt_4_all 25 Sep 2023 21:41 UTC
  1 point
  0
  Parent
  Got it, thanks!