Real CSI masked-autoencoder behind feature `tch-backend` (ADR-027 §2.0):
- CsiMae: dual-stream per-token amp+phase embed → fuse → residual-MLP encoder
over the visible tokens → flatten-to-latent bottleneck → learned per-position
query + broadcast latent → residual-MLP decoder → dec_amp_head / dec_ph_head
→ index_select the masked positions. (MLP-based v0; self-attention transformer
blocks are iter 3.)
- CsiMae::reconstruction_loss(pred_amp, pred_phase, tgt_amp, tgt_phase, phase_w)
= MSE(amp) + phase_w * MSE(phase).
- MaeBatch::from_windows — partition computed once from window 0 and reused
across the batch (the bottleneck fixes n_tokens), ndarray → tch conversion.
- pretrain_step(model, opt, batch) -> f64 — one Adam step, returns the loss.
- src/bin/pretrain_mae.rs — synthetic-data pre-train driver (required-features
= ["tch-backend"]); clap args for epochs/batch/samples/lr/mask-ratio/save.
- #[cfg(feature="tch-backend")] smoke test: loss halves when overfitting one
batch over 60 steps; also asserts model.n_visible/n_masked match
mask_csi_window's clamping.
v0 limits (documented in the module): fixed n_tokens; batch-shared masking;
MSE on unwrapped phase (vs a circular loss). The dev box has no LibTorch, so the
tch path is CI-verified (`--features tch-backend`), not locally. The default
`cargo test -p wifi-densepose-train --no-default-features` stays green (121 lib
tests) — the model module and the bin are both feature-gated.
Co-Authored-By: claude-flow <ruv@ruv.net>
csi_mae::mask_csi_window now dispatches on MaskStrategy:
- Random: uniform Fisher–Yates (as before).
- InfoGuided: CIG-MAE-style — preferentially mask high-information tokens.
A token's "information" = variance of its amplitude values +
variance of its phase values (token_information()); near-constant
tokens are trivially in-painted so masking them teaches less.
Selection is weighted-without-replacement (Efraimidis–Spirakis:
key_i = u_i^(1/w_i), ranked by ln(u_i)/w_i) — exact, and
deterministic given `seed` (the u_i come from SplitMix64).
Replaces the iteration-1 "InfoGuided falls back to Random with a warning" stub.
+3 unit tests (info-guided skews ≥7.5/10 toward high-info tokens; deterministic
in seed; token_information ≈ 0 for constant tokens). `cargo test -p
wifi-densepose-train --no-default-features` → 121 lib tests pass.
Still to do (iter 2b, next loop tick): the real csi_mae::model (tch encoder/
decoder + reconstruction_loss + pretrain_step), bin/pretrain_mae.rs, a gated
"loss decreases" smoke test.
Co-Authored-By: claude-flow <ruv@ruv.net>
New `wifi-densepose-train::csi_mae` module (ADR-027 §2.0):
- MaeConfig (+ validate), MaskStrategy {Random, InfoGuided}
- TokenLayout — flattens a [T,tx,rx,sub] CSI window to [N=T*tx*rx, sub] tokens
(the same layout model.rs::ModalityTranslator consumes)
- mask_csi_window — deterministic visible/masked token partition + amplitude &
phase reconstruction targets; reproducible via a tiny inline SplitMix64 PRNG
(no extra dependency); clamps so both partitions are non-empty
- reassemble_tokens — round-trips encoder-visible + decoder-predicted tokens
back to a full [N, sub] grid (for reconstruction eval/viz)
- model submodule (gated behind `tch-backend`): v0 skeleton — the
encoder/decoder networks, reconstruction loss, and pretrain_step land in
iteration 2 (transformer blocks, per-sample masking, info-guided masking,
a `pretrain-mae` bin)
8 new unit tests; builds and tests green under
`cargo test -p wifi-densepose-train --no-default-features` (118 lib tests pass).
The tch-gated `model` submodule is not exercised by the default workspace test
job — compile-checking it needs a LibTorch toolchain.
Co-Authored-By: claude-flow <ruv@ruv.net>
The Rust port lived two directories deep (rust-port/wifi-densepose-rs/)
without any sibling under rust-port/ that warranted the extra level.
Move the whole workspace up to v2/ to match v1/ (Python) at the same
depth and shorten every cd / build command across the repo.
git mv preserves history for all tracked files. 60 files updated for
path references (CI workflows, ADRs, docs, scripts, READMEs, internal
.claude-flow state). Two manual fixes for relative-cd paths in
CLAUDE.md and ADR-043 that became wrong after the depth change
(cd ../.. → cd ..).
Validated:
- cargo check --workspace --no-default-features → clean (after target/
nuke; the gitignored target/ was carried by the OS rename and had
hard-coded old paths in build scripts)
- cargo test --workspace --no-default-features → 1,539 passed, 0 failed,
8 ignored (same totals as pre-rename)
- ESP32-S3 on COM7 → still streaming live CSI (cb #40300, RSSI -64 dBm)
After-merge follow-up: contributors should `rm -rf v2/target` once and
let cargo regenerate from the new path.
ruv