"""Next Ask inference: predict digits and choose the next examples to label. Usage: python inference.py --selftest from inference import load, predict, next_queries """ import json import os import sys HERE = os.path.dirname(os.path.abspath(__file__)) sys.path.insert(0, HERE) import numpy as np # noqa: E402 import ask_core as ak # noqa: E402 from nlrt import load_model # noqa: E402 def load(path=os.path.join(HERE, "model.json")): return load_model(path) def predict(spec, pixels): """pixels: (N, 64) ints 0..16 (8x8 digits). Returns class probabilities (N, 10).""" return ak.proba(spec["tensors"]["W"], ak.features(pixels)) def next_queries(spec, pool_pixels, labeled, labels, strategy="hybrid", batch=5, seed=0): """Retrain on the labeled pool items and return the next `batch` pool indices to label.""" X = ak.features(pool_pixels) W = ak.train(X[labeled], np.asarray(labels), **spec["meta"]["hyper"]) return ak.select(strategy, W, X, labeled, batch, seed) def _selftest(): spec = load() with open(os.path.join(HERE, "selftest.json"), encoding="utf-8") as f: t = json.load(f) P = predict(spec, t["pixels"]) diff = float(np.abs(P - np.asarray(t["expected_proba"])).max()) picks = next_queries(spec, t["pool_pixels"], t["labeled"], t["labels"], t["strategy"], t["batch"], t["seed"]) ok = diff < 1e-6 and picks == t["expected_picks"] print(json.dumps({"passed": bool(ok), "max_abs_diff": diff, "picks": picks, "expected_picks": t["expected_picks"], "n_samples": len(t["pixels"])})) return ok if __name__ == "__main__": if "--selftest" in sys.argv: sys.exit(0 if _selftest() else 1) print(__doc__)