5ffa38fca42bb3fbee030f4b80264e0f4910a49c,torch_geometric/datasets/utils/tu_format.py,,read_adj,#Any#Any#,15

Before Change

    new_index = index.new(index.size()).copy_(index)

    indicator = read_file(dir, prefix, "graph_indicator").long() - 1
    bincount = torch.from_numpy(np.bincount(indicator.numpy()))

    index_slice = index.new(bincount.size(0) + 1)
    index_slice[0] = 0
    curr_graph = indicator[0]
    for i in range(index.size(1)):
        row = index[0, i]
        if indicator[row] > curr_graph:
            new_index[:, i:] -= bincount[curr_graph]
            curr_graph += 1
            index_slice[curr_graph] = i
    index_slice[-1] = index.size(1)
    index = new_index

After Change

    row, col = index
    row, perm = row.sort()
    col = col[perm]
    index = torch.stack([row, col], dim=0)
    indicator = read_file(dir, prefix, "graph_indicator").long() - 1

    index_slice = index.new(indicator.max() + 2)
    index_slice[0] = 0
    index_slice[-1] = index.size(1)
    curr_graph = indicator[0]

    for i in range(index.size(1)):
        row = index[0, i]
        if indicator[row] > curr_graph:
            index[:, index_slice[curr_graph]:i] -= index[:, index_slice[curr_graph]:i].min()
            curr_graph += 1
            index_slice[curr_graph] = i
    index[:, index_slice[curr_graph]:] -= index[:, index_slice[curr_graph]:].min()

    return index, index_slice

def read_slice(dir, prefix):

Instances