Files
Dragan Spiridonov 3ed43e9a2f fix(python): ship SOTA bindings in release wheels; make parity tests arch-portable
Three coupled defects, all found by building the wheel and running the real
suite on Apple Silicon rather than trusting green CI.

1. THE P6 SOTA BINDINGS NEVER REACHED USERS (release blocker).
   `pip-release.yml`'s cibuildwheel built the DEFAULT feature set, so published
   wheels contained none of aether/mat/meridian. `pip install
   wifi-densepose[aether]` then raised ImportError — and the extra is empty, so
   its own error message ("install the [aether] extra") sent users in a circle.
   A pip extra cannot enable a Rust cargo feature on an already-built wheel, so
   the only way P6 reaches PyPI is to compile it in.

   Fix: the RELEASE build opts in via `MATURIN_PEP517_ARGS="--features sota"`
   (per-platform, since CIBW_ENVIRONMENT_LINUX overrides CIBW_ENVIRONMENT).
   `default = []` in Cargo.toml, the RuView#1387-default-wheel-budget-config
   fix-marker, and the wheel-size-budget job are all left UNTOUCHED — they keep
   guarding the small base compile. Measured published wheel: 1.68 MiB, well
   under the ADR-117 §5.4 5 MiB budget. Proven: built via the exact PEP517 path,
   `import wifi_densepose.aether/mat/meridian` all succeed.

2. THE WHEEL SMOKE TEST COULD NOT SEE #1.
   CIBW_TEST_COMMAND only asserted `hello()` and PRINTED `__build_features__`.
   The one signal that would reveal missing bindings was dumped to stdout and
   ignored, so a featureless wheel published green. It now ASSERTS the p6
   features are present and imports the three modules. Proven red→green: fails
   on the old featureless wheel ("missing SOTA bindings: [...]"), passes on the
   fixed one.

3. THE AETHER PARITY TESTS WERE NOT PORTABLE ACROSS THE WHEEL MATRIX.
   `test_aether.py` and the native `aether_parity.rs`/`aether_weights_parity.rs`
   hashed the raw f32 embedding bytes (SHA-256) against a committed golden. The
   embedding is pure f32 with transcendental ops (ln/sqrt/cos) that are not
   bit-reproducible across CPUs/libm, so the hash only ever matched the one arch
   that generated it. This passed in python-ci (x86) but fails on the aarch64 /
   macOS-arm wheels this same project ships — latent until #1 is fixed and the
   bindings actually load. Every tolerance/behavioral assertion already passed;
   only the two byte-hash tests failed, which is the signature of a non-portable
   golden, not a logic bug.

   Fix: compare to a committed golden VECTOR within tolerance
   (atol=rtol=1e-4 — ~100x cross-arch f32 drift, ~100x under any real algorithm
   change), on both the Python and native sides against the SAME golden. Native
   ≈ golden and binding ≈ golden together prove binding ≈ native, portably.
   `.sha256` goldens replaced by `.json` vectors.

Also: the aether/mat/meridian import shims told users to `pip install
wifi-densepose[<x>]` on a missing feature — an empty extra that cannot help.
Corrected to name the real fix (rebuild with `--features <x>`); message tests
updated to assert the honest message and forbid the misleading one.

Verified on aarch64/macOS: full `python/tests/` suite 227 passed against a wheel
built with the new mechanism; `cargo check --tests --features aether` clean.
The native `.rs` parity tests were converted by inspection and cargo-checked but
not executed — they link against the PyO3 crate and no workflow runs them today
(a pre-existing gap; wiring `cd python && cargo test --features sota` into
python-ci would make the native anchor actually run).

Co-Authored-By: Ruflo & AQE
2026-07-24 10:45:00 +02:00

112 lines
4.3 KiB
Rust

//! ADR-185 §4.1 — AETHER parity: native-Rust reference half.
//!
//! Produces the golden 128-dim embedding by calling the canonical
//! `wifi-densepose-aether::embedding` code DIRECTLY (no PyO3), for the
//! committed `tests/golden/aether_input.json` fixture, and compares it to the
//! committed golden VECTOR `tests/golden/aether_embedding.json` within a
//! numerical tolerance.
//!
//! Why a vector + tolerance and not a SHA-256 of the f32 bytes: the embedding
//! is pure f32 and uses transcendental ops (ln/sqrt/cos), which are not
//! bit-reproducible across CPU architectures or libm implementations. A byte
//! hash only ever matched the one arch that generated it and failed on every
//! other wheel this project builds (aarch64, macOS-arm). The pytest half
//! (`tests/test_aether.py`) compares the Python binding to the SAME golden
//! within the same tolerance — native≈golden and binding≈golden together prove
//! binding≈native, portably.
//!
//! Regeneration (only when the Rust subsystem intentionally changes): delete
//! `tests/golden/aether_embedding.json` and re-run `cargo test --features aether`.
#![cfg(feature = "aether")]
use std::fs;
use std::path::PathBuf;
use wifi_densepose_aether::embedding::{EmbeddingConfig, EmbeddingExtractor};
use wifi_densepose_aether::graph_transformer::TransformerConfig;
/// Cross-architecture f32 parity tolerance; see the module docs and the
/// matching `PARITY_ATOL`/`PARITY_RTOL` in `tests/test_aether.py`.
const PARITY_ATOL: f32 = 1e-4;
const PARITY_RTOL: f32 = 1e-4;
/// Assert `embedding` matches the committed golden vector `<name>` within
/// tolerance, or (if the golden is absent) write it and fail asking for a re-run.
fn assert_matches_golden_vector(embedding: &[f32], name: &str) {
let path = golden_dir().join(name);
match fs::read_to_string(&path) {
Ok(raw) => {
let golden: Vec<f32> = serde_json::from_str(&raw)
.expect("parse golden vector json");
assert_eq!(embedding.len(), golden.len(), "{name}: length mismatch");
for (i, (&got, &want)) in embedding.iter().zip(&golden).enumerate() {
let tol = PARITY_ATOL + PARITY_RTOL * want.abs();
assert!(
(got - want).abs() <= tol,
"{name}: element {i} diverged beyond tolerance \
(got {got}, golden {want}, |Δ|={}) — a real regression, \
not cross-arch f32 drift",
(got - want).abs()
);
}
}
Err(_) => {
let json = serde_json::to_string(&embedding).expect("serialize golden");
fs::write(&path, &json).expect("write golden vector");
panic!("no committed golden {name}; wrote it. Re-run to verify parity.");
}
}
}
fn golden_dir() -> PathBuf {
PathBuf::from(env!("CARGO_MANIFEST_DIR"))
.join("tests")
.join("golden")
}
fn load_input() -> Vec<Vec<f32>> {
let raw = fs::read_to_string(golden_dir().join("aether_input.json"))
.expect("read aether_input.json fixture");
let rows: Vec<Vec<f64>> = serde_json::from_str(&raw).expect("parse aether_input.json");
rows.into_iter()
.map(|row| row.into_iter().map(|x| x as f32).collect())
.collect()
}
/// Build the extractor identically to the Python binding's default
/// construction: `AetherConfig()` + `EmbeddingExtractor(n_subcarriers=56, cfg)`.
fn embed_native(input: &[Vec<f32>]) -> Vec<f32> {
let e_config = EmbeddingConfig {
d_model: 64,
d_proj: 128,
temperature: 0.07,
normalize: true,
};
let t_config = TransformerConfig {
n_subcarriers: 56,
n_keypoints: 17,
d_model: 64,
n_heads: 4,
n_gnn_layers: 2,
};
let mut ext = EmbeddingExtractor::new(t_config, e_config);
ext.extract(input)
}
#[test]
fn native_embedding_is_128_dim_unit_norm() {
let emb = embed_native(&load_input());
assert_eq!(emb.len(), 128, "AETHER embedding must be 128-dim");
let norm: f32 = emb.iter().map(|x| x * x).sum::<f32>().sqrt();
assert!(
(norm - 1.0).abs() < 1e-4,
"embedding must be L2-normalized, got norm={norm}"
);
}
#[test]
fn native_embedding_matches_committed_golden() {
let emb = embed_native(&load_input());
assert_matches_golden_vector(&emb, "aether_embedding.json");
}