Fix Glorot/He uniform init docstrings to label the uniform bound, not sigma

[vineethsaivs]

Proposed changes

The glorot_uniform and he_uniform docstrings in python/mlx/nn/init.py
write the sampling bound with the symbol \sigma, but both initializers draw
from a symmetric uniform distribution, uniform(-a, a), so the value shown is
the bound a, not a standard deviation. \sigma conventionally denotes the
standard deviation, and for uniform(-a, a) the standard deviation is
a / sqrt(3):

• glorot_uniform: bound a = gain * sqrt(6 / (fan_in + fan_out)), so the
  actual standard deviation is gain * sqrt(2 / (fan_in + fan_out)), which is
  exactly the \sigma documented for glorot_normal.
• he_uniform: bound a = gain * sqrt(3 / fan), so the actual standard
  deviation is gain / sqrt(fan), which is exactly the \sigma documented for
  he_normal.

So the documented \sigma for the uniform initializers is off by a factor of
sqrt(3) from the real standard deviation. It is also inconsistent with the
surrounding prose, which already describes the quantity as a "range", and with
the normal initializers in the same file, where \sigma correctly denotes the
standard deviation. For comparison, PyTorch's kaiming_uniform_ docstring
labels the same quantity bound and samples U(-bound, bound).

This relabels the quantity as the bound a and shows the sampling interval
[-a, a]. It is a documentation-only change with no effect on behavior; the
sampled values are unchanged.

Checklist

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes (docstring text only; verified it compiles and does not change black formatting)
• I have added tests that prove my fix is effective or that my feature works (documentation-only change; no code path is affected)
• I have updated the necessary documentation (if needed)

[Pablosinyores]

Confirmed correct against the implementation. Both glorot_uniform and he_uniform compute limit = gain * math.sqrt(6.0 / (fan_in + fan_out)) (resp. sqrt(3.0 / fan)) and then mx.random.uniform(-limit, limit, ...) — so that quantity is the half-width of the uniform interval, not a standard deviation. Labeling it \sigma overstates the spread by a factor of √3 (for U[-a, a] the actual std is a/√3), so the relabel is right. The explicit [-a, a] interval line you added makes the meaning unambiguous, which is the better framing.

Worth noting the normal siblings (glorot_normal / he_normal) correctly keep \sigma alongside mx.random.normal(scale=std) — so this reads like the \sigma label was carried over from the normal variants into the uniform ones by mistake.

One small thought: the implementation names this variable limit, so using \text{limit} as the symbol (as your #3835 does) ties the formula straight to the source. a with the explicit interval is also perfectly clear and arguably more conventional — either works.

Heads up that this overlaps with your own #3835 — same two functions, same lines — so they'll conflict and one probably wants closing. This is the more complete of the two (the interval prose is the useful part), so if you keep one, I'd keep this.

[vineethsaivs]

Thanks for the careful review. Agreed the \sigma label looks carried over from the normal siblings. I have consolidated the two PRs into this one: it keeps the explicit [-limit, limit] interval wording and now uses \text{limit} as the symbol so the formula matches the code's variable name, as you suggested. Closing #3835 in favor of this PR.