torch_qr_fail.py

torch_qr_fail.py

import torch
import torch.nn as nn
import multiprocessing as mp

def get_flattened(tensor):
    """ This is just the first part of orthogonal initialization.
    Taken directly from https://github.com/pytorch/pytorch/blob/master/torch/nn/init.py
    """
    rows = tensor.size(0)
    cols = tensor[0].numel()
    flattened = torch.Tensor(rows, cols).normal_(0, 1)
    return flattened

def get_mat():
    """ This is what we send when during the initialization """
    wc_dim = 200
    mat = nn.utils.weight_norm(nn.Linear(wc_dim, wc_dim, bias=False), name="weight")
    tensor = mat.weight.data
    return tensor
 
def exp():
    tensor = get_mat()
    flattened = get_flattened(tensor)
    # It should fail if being called
    # in a child process.
    return torch.qr(flattened)

if __name__=="__main__":
    # First call
    result = exp()
    assert result is not None, "Oops, it should have not failed here"
    # Now create child process.
    p = mp.Process(target=exp)
    p.start()
    p.join()
    assert p.exitcode != 0, "Child process call to qr did not fail"
    assert p.exitcode == -11, "The error is not a segmentation fault %s" % p.exitcode
    print("Child process call to QR failed as expected")