ruvnet--RuView

mirror of https://github.com/ruvnet/RuView synced 2026-06-29 13:33:19 +00:00

Files

T

ruv 7fb3e33557 fix(train): rapid_adapt real finite-difference gradients, not a fake step (ADR-155 §Tier-1.3)

contrastive_step/entropy_step wrote a fake gradient (grad += v*0.01) unrelated
to the stated objective, so any "TTA improves the metric" was unsupported. The
*_loss functions are now pure evaluators of the real objective; adapt() descends
them with a central finite-difference gradient of that exact loss, so "the
adaptation loss decreases" is now a real, reproducible measurement.

Honest scope caveat (documented): this minimizes a self-supervised proxy over a
LoRA bottleneck on raw CSI; it is NOT wired to the pose model and there is NO
measured end-to-end PCK gain on WiFi pose from this path.

Tests: contrastive_loss_decreases, entropy_loss_decreases (real gradient steps
don't increase the loss), reported_loss_is_the_real_objective_not_a_placeholder.

Co-Authored-By: claude-flow <ruv@ruv.net>

2026-06-11 19:57:15 -04:00

.cargo

fix(security): audit — fix RUSTSEC vulns, clippy warnings, dead code (#769 )

2026-05-23 05:36:13 -04:00

.claude-flow

chore(repo): rename rust-port/wifi-densepose-rs → v2/ (flatten to one level) (#427 )

2026-04-25 21:28:13 -04:00

crates

fix(train): rapid_adapt real finite-difference gradients, not a fake step (ADR-155 §Tier-1.3)

2026-06-11 19:57:15 -04:00

data

chore(repo): rename rust-port/wifi-densepose-rs → v2/ (flatten to one level) (#427 )