tensorflow/models · error · ValueError

src_num_blocks must be equal to tgt_num_blocks.

Error message

src_num_blocks must be equal to tgt_num_blocks.

What it means

Error "src_num_blocks must be equal to tgt_num_blocks." thrown in tensorflow/models.

Source

Thrown at official/nlp/modeling/layers/block_sparse_attention.py:276

          training=training,
      )
      # pytype: enable=attribute-error
    # src_num_blocks and tgt_num_blocks are the number of blocks in the source
    # and target. Care should be taken to ensure that the number of blocks in
    # the source and target are the same.
    if self._query_shape[-2] % self._src_block_size != 0:
      raise ValueError(
          "query_shape[-2] must be divisible by src_block_size."
      )
    if self._key_shape[-2] % self._tgt_block_size != 0:
      raise ValueError(
          "key_shape[-2] must be divisible by tgt_block_size."
      )
    src_num_blocks = self._query_shape[-2] // self._src_block_size
    tgt_num_blocks = self._key_shape[-2] // self._tgt_block_size

    if src_num_blocks != tgt_num_blocks and tgt_num_blocks != 1:
      raise ValueError(
          "src_num_blocks must be equal to tgt_num_blocks."
      )
    # Convert the query/key/value into blocks to perform block diagonal
    # attention.
    query_blocks = tf.reshape(query, [
        -1,
        self._num_heads,
        src_num_blocks,
        self._src_block_size,
        self._key_dim,
    ])
    if tgt_num_blocks != 1 and self._num_kv_heads != 1:
      key_blocks = tf.reshape(key, [
          -1,
          self._num_heads,
          tgt_num_blocks,
          self._tgt_block_size,
          self._key_dim,

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/nlp/modeling/layers/block_sparse_attention.py:276 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/2b75f51bb9be8758. Report an issue: GitHub.