python/dglke/dataloader/sampler.py [111:145]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        rel_dict[r] = r_parts

    for i, edge_cnt in enumerate(edge_cnts):
        print('part {} has {} edges and {} relations'.format(i, edge_cnt, rel_cnts[i]))
    print('{}/{} duplicated relation across partitions'.format(num_cross_part, len(cnts)))

    parts = []
    for i in range(n):
        parts.append([])
        rel_parts[i] = np.array(rel_parts[i])

    for i, r in enumerate(rels):
        r_part = rel_dict[r][0]
        part_idx = r_part[0]
        cnt = r_part[1]
        parts[part_idx].append(i)
        cnt -= 1
        if cnt == 0:
            rel_dict[r].pop(0)
        else:
            rel_dict[r][0][1] = cnt

    for i, part in enumerate(parts):
        parts[i] = np.array(part, dtype=np.int64)
    shuffle_idx = np.concatenate(parts)
    heads[:] = heads[shuffle_idx]
    rels[:] = rels[shuffle_idx]
    tails[:] = tails[shuffle_idx]
    if has_importance:
        e_impts[:] = e_impts[shuffle_idx]

    off = 0
    for i, part in enumerate(parts):
        parts[i] = np.arange(off, off + len(part))
        off += len(part)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



python/dglke/dataloader/sampler.py [218:252]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        rel_dict[r] = r_parts

    for i, edge_cnt in enumerate(edge_cnts):
        print('part {} has {} edges and {} relations'.format(i, edge_cnt, rel_cnts[i]))
    print('{}/{} duplicated relation across partitions'.format(num_cross_part, len(cnts)))

    parts = []
    for i in range(n):
        parts.append([])
        rel_parts[i] = np.array(rel_parts[i])

    for i, r in enumerate(rels):
        r_part = rel_dict[r][0]
        part_idx = r_part[0]
        cnt = r_part[1]
        parts[part_idx].append(i)
        cnt -= 1
        if cnt == 0:
            rel_dict[r].pop(0)
        else:
            rel_dict[r][0][1] = cnt

    for i, part in enumerate(parts):
        parts[i] = np.array(part, dtype=np.int64)
    shuffle_idx = np.concatenate(parts)
    heads[:] = heads[shuffle_idx]
    rels[:] = rels[shuffle_idx]
    tails[:] = tails[shuffle_idx]
    if has_importance:
        e_impts[:] = e_impts[shuffle_idx]

    off = 0
    for i, part in enumerate(parts):
        parts[i] = np.arange(off, off + len(part))
        off += len(part)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -