CoLA Sparse operator is not producing the same result as torch coo sparse

[alxhosse]

I'm currently using CoLA version 0.0.6 (which seems uses COO format for Sparse operator) and PyTorch version 2.3.1, and I’ve encountered an issue when working with sparse matrices. I’ve noticed that I get inconsistent results when comparing the dense representations of sparse matrices created using torch.sparse_coo_tensor and CoLA's Sparse operator. Here's a simplified version of my code:

from cola.ops import Sparse
import torch
N = 4
nnz = 4 
row = torch.tensor([1, 3, 2, 2 ])
col = torch.tensor([1, 3, 2, 1 ])
data = torch.rand(nnz, dtype=torch.float64)
indices = torch.stack([row, col])
coo_tensor_torch = torch.sparse_coo_tensor(indices, data, size=[N, N])
coo_op_cola = Sparse(data, row, col, (N,N))
torch.isclose(coo_op_cola.to_dense(),coo_tensor_torch.to_dense()).all() # This outputs False

The issue arises when I compare the dense matrix representations of the CoLA sparse matrix and the PyTorch sparse matrix. The result of torch.isclose(...).all() is False, suggesting a mismatch between the two.

However, when I sort the row and column indices before creating the sparse matrix, the results match, as shown in this modified version:

from cola.ops import Sparse
import torch
N = 4
nnz = 4 
row = torch.sort(torch.tensor([1, 3, 2, 2 ]))[0]
col = torch.sort(torch.tensor([1, 3, 2, 1 ]))[0]
data = torch.rand(nnz, dtype=torch.float64)
indices = torch.stack([row, col])
coo_tensor_torch = torch.sparse_coo_tensor(indices, data, size=[N, N])
coo_op_cola = Sparse(data, row, col, (N,N))
torch.isclose(coo_op_cola.to_dense(),coo_tensor_torch.to_dense()).all()

I'm not sure why this is happening. I'm not sure if this is a bug or something I haven't understoond.