For example, see this snippet from an old Google repo: https://github.com/google/flaxformer/blob/ee62754ebe5a5eeb11...
I certainly don't think it will help at all with stability. Things like Q/K layernorm are better tricks for softmax stability when scaling: https://arxiv.org/pdf/2302.05442.pdf
How would you have known if the trick actually reduces the outliers in the weights? Even if the transformer quality does not improve overall, having less outliers as a result is very beneficial for more accurate quantization of the data
The "how" is pretty straightforward.
>> because no one has yet looked at whether the trick helps reducing outliers in very large models
Given a softmax version doing exactly as the blog post says is baked into a google library (see this thread), and you can set it as a parameter in a pytorch model (see this thread), this claim seems off. "Let's try X, oh, X doesn't do much, let's not write a paper about it" is extremely common for many X.
Agree - the "how" is straightforward
Technically softmax is not implemented as presented but through exp(x_i-max(x)), and summing over it in the denom. But maybe I am missing something.
Furthermore, the residuals are used exactly because the networks cant learn the identity function; but they can learn zero; at which point the residual is `f(x): x+g(x)` with being `g:x ~> 0` (ie approximately 0).
It is also the case that `f(x): x+g(x)` makes it easier for gradients to flow through.
Regardless of numerical stability tricks (e.g. exp(x_i-max(x))), you are still simply normalizing the logits such that the probabilities sum to 1.
The blog adds an additional hidden logit (equal to 0) to allow for softmax(x) = 0 when x -> -inf.
Enough weights don't fall under that "nearly" that we require more bits per weight to cover those edge cases. If we were able to delete the "nearly" we would need fewer bits (smaller models).