hankcs/HanLP · error · ValueError

You have to specify either input_ids or inputs_embeds

Error message

You have to specify either input_ids or inputs_embeds

What it means

CategoricalAccuracy requires top_k >= 1 since it computes the top-k predictions; top_k=0 or negative values are meaningless and raise ValueError immediately in __init__.

Source

Thrown at hanlp/components/amr/amrbart/model_interface/modeling_bart.py:796

            return_dict (`bool`, *optional*):
                Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        # retrieve input_ids and inputs_embeds
        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            input_shape = input_ids.size()
            input_ids = input_ids.view(-1, input_shape[-1])
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale

        embed_pos = self.embed_positions(input_shape)

        hidden_states = inputs_embeds + embed_pos
        hidden_states = self.layernorm_embedding(hidden_states)
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)

        # expand attention_mask
        if attention_mask is not None:
            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
            attention_mask = _expand_mask(attention_mask, inputs_embeds.dtype)

        encoder_states = () if output_hidden_states else None
        all_attentions = () if output_attentions else None

View on GitHub (pinned to ddb1299bdd)

Solutions

1. Set top_k to a positive int (1 for exact match, k<=num_classes for top-k)
2. Validate top_k > 0 in config-loading code before constructing the metric
3. Default to top_k=1 by omitting the argument

Example fix

# before
metric = CategoricalAccuracy(top_k=0)
# after
metric = CategoricalAccuracy(top_k=1)

assert isinstance(top_k, int) and top_k >= 1, f'top_k must be >= 1, got {top_k}'

def valid_top_k(k) -> bool:
    return isinstance(k, int) and k >= 1

try:
    m = CategoricalAccuracy(top_k=top_k)
except ValueError:
    m = CategoricalAccuracy(top_k=1)

When it happens

Trigger: Constructing CategoricalAccuracy(top_k=0) or CategoricalAccuracy(top_k=-1), often via a miscomputed config value.

Common situations: Config generation code computing top_k arithmetically (e.g. num_classes - num_classes); JSON/YAML typo 0 or -1; passing None-like defaults that become 0.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• You cannot specify both input_ids and inputs_embeds at the s
• The head_mask should be specified for {len(self.layers)} lay
• You cannot specify both decoder_input_ids and decoder_inputs
• You have to specify either decoder_input_ids or decoder_inpu
• The `{mask_name}` should be specified for {len(self.layers)}

AI-assisted analysis of hankcs/HanLP@ddb1299bdd (2026-08-27). Data as JSON: /api/errors/9e63b4fa9f3d7575. Report an issue: GitHub.