Saving 3D array into `.npy` file for large dimension gets corrupted output

[sahilhassanzee]

I'm trying to put a 3D array of dimension (10000, 160, 346) in an .npy file using the append mode as shown in the code below:

#include "cnpy.h"
#include <vector>
#include <cstdlib>
constexpr int N = 346*160;

int main(int argc, char* argv[]) {
    int ROWS = 10000;

    // Create a vector
    std::vector<double> vec (N,0);

    // Fill the vector with data
    for(int i=0; i < N; i++) {
        vec[i] = i;
    }

    // Append the vector ROWS number of times into an npz file
    for (int t=0; t<ROWS; t++){
        cnpy::npy_save("sample.npy", &vec[0], {1,160,346}, "a");
    }

    return 0;
}

If I now try to open this file using the following Python code:

import numpy as np
x = np.load('sample.npy')

I get the following error:

ValueError: cannot reshape array of size 553599998 into shape (10000,160,346)

So, the size of array in the saved sample.npy is 2 Bytes smaller than what's expected. I have observed arrays upto 16 Bytes off from the expected size for larger size. Could you help resolve this issue?

[sahilhassanzee]

By the way, my numpy version is 2.1.3