hankcs/HanLP · error · ValueError

The `{mask_name}` should be specified for {len(self.layers)}

Error message

The `{mask_name}` should be specified for {len(self.layers)} layers, but it is for {head_mask.size()[0]}.

What it means

When a mask is supplied to the elementwise accuracy call, it must match predictions.shape exactly so the metric knows which cells to count. A mask with different shape (e.g. (B,) for (B, C) predictions) raises ValueError.

Source

Thrown at hanlp/components/amr/amrbart/model_interface/modeling_bart.py:1056

        # embed positions
        positions = self.embed_positions(input_shape, past_key_values_length)

        hidden_states = inputs_embeds + positions
        hidden_states = self.layernorm_embedding(hidden_states)

        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        all_cross_attentions = () if (output_attentions and encoder_hidden_states is not None) else None
        next_decoder_cache = () if use_cache else None

        # check if head_mask/cross_attn_head_mask has a correct number of layers specified if desired
        for attn_mask, mask_name in zip([head_mask, cross_attn_head_mask], ["head_mask", "cross_attn_head_mask"]):
            if attn_mask is not None:
                if attn_mask.size()[0] != (len(self.layers)):
                    raise ValueError(
                        f"The `{mask_name}` should be specified for {len(self.layers)} layers, but it is for"
                        f" {head_mask.size()[0]}."
                    )

        for idx, decoder_layer in enumerate(self.layers):
            # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
            if output_hidden_states:
                all_hidden_states += (hidden_states,)
            dropout_probability = random.uniform(0, 1)
            if self.training and (dropout_probability < self.layerdrop):
                continue

            past_key_value = past_key_values[idx] if past_key_values is not None else None

            if self.gradient_checkpointing and self.training:

                if use_cache:
                    logger.warning(

View on GitHub (pinned to ddb1299bdd)

Solutions

1. Expand the mask to predictions' shape: mask.unsqueeze(-1).expand_as(predictions) or a Boolean multi-hot mask
2. Recompute the mask after any reshape of predictions
3. Validate mask.shape == predictions.shape before the call

Example fix

# before
metric(predictions, gold, mask)  # mask (B,)
# after
metric(predictions, gold, mask.unsqueeze(-1).expand_as(predictions))

assert mask is None or mask.size() == predictions.size(), (mask.shape if mask is not None else None, predictions.shape)

import torch
def mask_matches(mask: torch.Tensor, predictions: torch.Tensor) -> bool:
    return mask.size() == predictions.size()

try:
    metric(predictions, gold, mask)
except ValueError as e:
    if 'mask' in str(e) and mask is not None:
        metric(predictions, gold, mask.unsqueeze(-1).expand_as(predictions))
    else:
        raise

When it happens

Trigger: Passing mask of shape (B,) or (B, T) alongside (B, C) / (B, T, C) predictions to the same-shape __call__ variant.

Common situations: Reusing a length-based 1-D mask from seq labeling in a multi-label metric; mask computed before a view/reshape of predictions; padding mask broadcasting assumptions.

Related errors

• The head_mask should be specified for {len(self.layers)} lay
• You have to specify either decoder_input_ids or decoder_inpu
• the first two dimensions of emissions and mask must match, g
• Unsupported argument type: {item}
• Attention weights should be of size {(bsz * self.num_heads,

AI-assisted analysis of hankcs/HanLP@ddb1299bdd (2026-08-27). Data as JSON: /api/errors/ab8d40889becef8d. Report an issue: GitHub.