The dataset conversion to ladder format about vecalign HOT 2 OPEN

def reformat(ladder_file, src_len, tgt_len):
    alignments = []
    current_alignment = ([], [])
    prev_a1, prev_a2 = None, None

    for line in open(ladder_file, 'r', encoding='utf-8'):
        fields = line.strip().split('\t')
        a1, a2 = int(fields[0]), int(fields[1])
            
        if a1 != prev_a1 and a2 != prev_a2 and current_alignment != ([], []):
            alignments.append(current_alignment)
            current_alignment = ([], [])
            
        current_alignment[0].append(a1)
        current_alignment[1].append(a2)
        prev_a1, prev_a2 = a1, a2
    
    alignments2 = []
    xx, yy = [], []
    for a1, a2 in alignments:
        x1 = sorted(list(set(a1)))
        x2 = sorted(list(set(a2)))
        alignments2.append((x1, x2))  # tuple of lists
        xx.extend(x1)
        yy.extend(x2)
    
    # add deletions/insertions (*not* in order) 
    xx, yy = set(xx), set(yy)
    for x in range(src_len):
        if x not in xx:
            alignments2.append(([x, ], []))
    for y in range(tgt_len):
        if y not in yy:
            alignments2.append(([], [y, ]))

    return alignments2