code wiki / _hdl_build / nx_builder_authors_tensor_grad.nx
nx_builder_authors_tensor_grad.nx
buildroot/runtime/_hdl_build/nx_builder_authors_tensor_grad.nx
about
nx_builder_authors_tensor_grad.nx -- the BUILDER authors T6 TENSOR AUTOGRAD (training-substrate
rung 2): array-valued tape over a deterministic bump arena + the four tutored backward identities
(matvec outer-product/transpose, addvec, relu-per-cell, MSE). Claude tutored the identities
(knowledge/specs/2026-06-09-tutoring-training-substrate-rung2-tensor.md); the Builder authors;
the Engineer gates via nx_sov_build_run. Gates: 9-param MLP gradcheck / affine-map recovery from
zero init (convex by design; nonconvex init = flagged next rung) / bit-exact two-run training.
license_tier: ORIGINAL
dependencies 2 imports · 0 importers
imports: nx_module_author.nxnx_syscalls.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| none |
functions
| 10 | func _b(s: *u8) -> i64 { var n: i64=0; while s[n]!=(0 as u8){n=n+1} sys_write(1,s,n); return 0 } |
| 11 | func main() -> i64 |