🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

SGD vs Adam💡 I punish x10 for distance to mouse in the loss function, but Adam's gradient normalisation eliminates the effect. #MLX #SwiftUI #MachineLearning

The discourse over the Apple paper about how LLMs can't reason still shows that there is an unsatiable thirst over negativity in anything deep learning related. You can get tens of thousand of likes claiming bullshit in a game of telephone because people want to believe in it.

Enlightment

A lot of linear RNNs and test time training models are now using forgetting gates, but we don't really do that with standard optimizers like Adam or Muon yet..

My brain is simulating you rn, but cloth simulations are computationally challenging so we'll have to make do without your clothes in it

My timeline blessing me with the hardest girls to fix

Played "Peak" with friends, it was nice, but there's a complete lack of pausing the game in any way, even singleplayer offline. Given that good runs last hours and wasting time penalizes you, it's a big issue. Don't recommend until they fix this.

I just looked up what mewing is about and it turns out I've been naturally doing it my whole life, ama

"How to get strong to do the planche" and it's a short guy below 60kg in the thumbnail

Empirical Fisher (aka estimating Hessian using training labels instead of sampling from model outputs) is one of the biggest scams surrounding neural networks.