tensorflow/models · error

Found unexpected kwargs for `grad`:

Error message

Found unexpected kwargs for `grad`: 

What it means

Error "Found unexpected kwargs for `grad`: " thrown in tensorflow/models.

Source

Thrown at official/projects/bigbird/recompute_grad.py:217

      # Use the cached context in the recomputation phase.
      with parent_context.children.popleft()._replace(
          is_recomputing=True) as context:
        result = f(*args, **kwargs)
    else:
      with RecomputeContext(
          is_recomputing=False,
          seed=_make_seed_if_none(seed),
          children=collections.deque()) as context:
        result = f(*args, **kwargs)
        # In the forward pass, build up a tree of recomputation contexts.
        if parent_context is not None and not parent_context.is_recomputing:
          parent_context.children.append(context)

    def grad(*dresult, **grad_kwargs):
      """Gradient function calculation for inner function."""
      variables = grad_kwargs.pop('variables', None)
      if grad_kwargs:
        raise ValueError('Found unexpected kwargs for `grad`: ',
                         list(grad_kwargs.keys()))
      inputs, seed = list(args), context.seed
      if _in_xla_context():
        inputs = _force_data_dependency(
            tf.nest.flatten(dresult), inputs + [seed])
        seed = inputs.pop()
      with tf.GradientTape() as tape:
        tape.watch(inputs)
        if variables is not None:
          tape.watch(variables)
        with tf.control_dependencies(dresult):
          with context._replace(is_recomputing=True, seed=seed):
            result = f(*inputs, **kwargs)
      kw_vars = []
      if variables is not None:
        kw_vars = list(variables)
      grads = tape.gradient(
          result, list(inputs) + kw_vars, output_gradients=dresult)

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/bigbird/recompute_grad.py:217 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/81f445800423baec. Report an issue: GitHub.