tensorflow/models · error · ValueError

Unknown value {shape_for_attn} for shape_for_attention.

Error message

Unknown value {shape_for_attn} for shape_for_attention.

What it means

Error "Unknown value {shape_for_attn} for shape_for_attention." thrown in tensorflow/models.

Source

Thrown at official/projects/perceiver/modeling/layers/utils.py:110

    A `dict` mapping `TransformerEncoderBlock` arguments.

  References:
    [Perceiver: General Perception with Iterative
    Attention](https://arxiv.org/abs/2103.03206)
    (https://github.com/deepmind/deepmind-research/blob/master/perceiver/perceiver.py)
    (https://github.com/tensorflow/models/blob/871c4e0a393ef4385534bee55354a5df8aa1ccf4/official/nlp/modeling/layers/transformer_encoder_block.py)
  """
  inputs_q_shape = input_shape[0]
  inputs_kv_shape = input_shape[1]

  output_last_dim = inputs_q_shape[-1]

  if shape_for_attn == "q":
    f_qk_last_dim = inputs_q_shape[-1]
  elif shape_for_attn == "kv":
    f_qk_last_dim = inputs_kv_shape[-1]
  else:
    raise ValueError(f"Unknown value {shape_for_attn} for "
                     "shape_for_attention.")

  f_v_last_dim = None
  if qk_last_dim is not None:
    f_qk_last_dim = qk_last_dim
  if v_last_dim is not None:
    f_v_last_dim = v_last_dim

  return _build_transformer_encoder_block_args(
      input_shape=input_shape,
      widening_factor=widening_factor,
      dropout_prob=dropout_prob,
      dropout_attn_prob=dropout_attn_prob,
      num_heads=num_heads,
      att_init_scale=att_init_scale,
      dense_init_scale=dense_init_scale,
      use_query_residual=use_query_residual,
      norm_epsilon=norm_epsilon,

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/perceiver/modeling/layers/utils.py:110 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tensorflow/models@e006f5f0d5 (2026-08-24). Data as JSON: /api/errors/4b4fc1ad1e9a3faa. Report an issue: GitHub.