hdc_data_shuffle_v2.py

import numpy as np
from numpy.lib.format import open_memmap

# === Input paths ===
X_in = "datasets/hdc_pairs_200_shot/X_pairs_combined.npy"
Y_in = "datasets/hdc_pairs_200_shot/Y_pairs_combined.npy"
O_in = "datasets/hdc_pairs_200_shot/O_pairs_combined.npy"

# === Output paths ===
X_out = "datasets/hdc_pairs_200_shot/X_pairs_shuffled.npy"
Y_out = "datasets/hdc_pairs_200_shot/Y_pairs_shuffled.npy"
O_out = "datasets/hdc_pairs_200_shot/O_pairs_shuffled.npy"

# === Load input data ===
X = np.load(X_in, mmap_mode="r")
Y = np.load(Y_in, mmap_mode="r")
O = np.load(O_in, mmap_mode="r")

N, W, F = X.shape
_, _, T = Y.shape
_, O_dim = O.shape

# === Create empty output arrays ===
X_shuf = open_memmap(X_out, mode="w+", dtype=np.float32, shape=(N, W, F))
Y_shuf = open_memmap(Y_out, mode="w+", dtype=np.float32, shape=(N, W, T))
O_shuf = open_memmap(O_out, mode="w+", dtype=np.float32, shape=(N, O_dim))

# === Generate and apply permutation ===
perm = np.random.default_rng().permutation(N)

for new_idx, old_idx in enumerate(perm):
    X_shuf[new_idx] = X[old_idx]
    Y_shuf[new_idx] = Y[old_idx]
    O_shuf[new_idx] = O[old_idx]

# === Flush to disk ===
X_shuf.flush()
Y_shuf.flush()
O_shuf.flush()

print(f"✅ Shuffled and saved:")
print(f"  → {X_out}")
print(f"  → {Y_out}")
print(f"  → {O_out}")