Skip to main content

Module muon

teeny_kernels::nn::optim

Module muon 

Source

Structs§

MuonFrobNormSq
Parallel Frobenius squared-norm via atomic reduction.
MuonNsStep
One Newton-Schulz step (in-place): X ← a·X + b·(T·X) or X ← a·X + b·(X·T).
MuonNsXtx
Gram matrix: T = X·Xᵀ (!TRANSPOSE) or T = Xᵀ·X (TRANSPOSE).
MuonUpdate
Parameter update: W -= lr · G_orth.

Functions§

muon_frob_norm_sq
Parallel Frobenius squared-norm via atomic reduction.
muon_ns_step
One Newton-Schulz step (in-place): X ← a·X + b·(T·X) or X ← a·X + b·(X·T).
muon_ns_xtx
Gram matrix: T = X·Xᵀ (!TRANSPOSE) or T = Xᵀ·X (TRANSPOSE).
muon_update
Parameter update: W -= lr · G_orth.