hankcs/HanLP · error · ValueError

Unsupported normalization method

Error message

Unsupported normalization method {normalize}

What it means

After selecting pretrained vectors, index_word2vec_with_vocab can normalize them; supported schemes are 'norm' (divide by row L2 norm), 'l2' (F.normalize p=2), and 'std' (divide by global std). Any other normalize string raises, so typos or unsupported schemes fail fast rather than silently skipping normalization.

Solutions

1. Use one of 'norm', 'l2', 'std'
2. Pass normalize=None (or omit) to skip normalization

Example fix

# before
emb = build_word2vec_with_vocab(p, vocab, normalize='minmax')  # error
# after
emb = build_word2vec_with_vocab(p, vocab, normalize='l2')

assert normalize in (None, 'norm', 'l2', 'std'), f'bad normalize: {normalize}'

def valid_normalize(norm) -> bool:
    return norm in (None, 'norm', 'l2', 'std')

When it happens

Trigger: Passing normalize='none', normalize=None is fine but 'max', 'minmax', 'unit' etc. raise the error.

Common situations: Assuming other normalization names from sklearn/gensim apply; typo in config keys.

Understand the failure class

Background: "Invalid value" and "allowed values are" config errors: what your library rejected and how to fix it — this error's family across 41 libraries.

Related errors

• Unsupported init
• embed_dim must be divisible by num_heads
• error
• Not supported.
• self.model.config.pad_token_id has to be defined.

AI-assisted analysis of hankcs/HanLP@ddb1299bdd (2026-08-27). Data as JSON: /api/errors/16bf8d892047a637. Report an issue: GitHub.

Appendix: source

Thrown at hanlp/layers/embeddings/util.py:74

    if unk_id_offset:
        unk_embeds = torch.zeros(unk_id_offset, pret_matrix.size(1))
        if init and init != 'zeros':
            if init == 'uniform':
                init = embedding_uniform
            else:
                raise ValueError(f'Unsupported init {init}')
            unk_embeds = init(unk_embeds)
        pret_matrix = torch.cat([pret_matrix, unk_embeds])
    ids = torch.LongTensor(ids)
    embedding = pret_matrix.index_select(0, ids)
    if normalize == 'norm':
        embedding /= (torch.norm(embedding, dim=1, keepdim=True) + 1e-12)
    elif normalize == 'l2':
        embedding = torch.nn.functional.normalize(embedding, p=2, dim=1)
    elif normalize == 'std':
        embedding /= torch.std(embedding)
    else:
        raise ValueError(f'Unsupported normalization method {normalize}')
    return embedding


def build_word2vec_with_vocab(embed: Union[str, int],
                              vocab: Vocab,
                              extend_vocab=True,
                              unk=None,
                              lowercase=False,
                              trainable=False,
                              init='zeros',
                              normalize=None) -> nn.Embedding:
    """Build word2vec embedding and a vocab.

    Args:
        embed:
        vocab: The vocabulary from training set.
        extend_vocab: Unlock vocabulary of training set to add those tokens in pretrained embedding file.
        unk: UNK token.

View on GitHub (pinned to ddb1299bdd)