apache/beam · error · ValueError

Invalid model update

Error message

Invalid model update: {key} appears in update, but not in the original configuration.

What it means

Raised by update_model_paths when an update references a key that was never part of the KeyedModelHandler's original per-key configuration. Updates can only replace paths for keys the handler already knows via its _key_to_id_map.

Solutions

Include only keys that exist in the handler's original configuration.
Fix key typos/mismatches between the initial setup and the update spec.
To add new cohorts, re-create the KeyedModelHandler with the full key set instead of using update_model_paths.

Example fix

// before
handler.update_model_paths([UpdateModelPath(keys=['k_new'], update_path='gs://b/m2', model_id='m2')])  # k_new never configured
// after
handler.update_model_paths([UpdateModelPath(keys=['k1'], update_path='gs://b/m2', model_id='m2')])  # k1 was in original config

Defensive patterns

Strategy: validation

Validate before calling

known = set(handler._key_to_id_map)
unknown = {k for u in updates for k in u.keys} - known
if unknown:
    raise ValueError(f'keys not in original config: {unknown}')

Prevention

Source update keys from the same configuration used to build the handler.
Re-create the handler when adding new cohorts instead of updating unknown keys.

When it happens

Trigger: Calling update_model_paths with an update whose keys include a key absent from the mapping used at handler construction (e.g. new cohort key 'k9' not in the original update_model_path setup).

Common situations: Adding a new model cohort via an update instead of rebuilding the handler; renamed/typo'd keys between initial setup and update; stale update specs generated from a different handler configuration.

Understand the failure class

Background: Record Not Found Errors: "not found", RecordNotFound, and "was not found" — what they mean and how to fix them — this error's family across 28 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/e23b356939c249de. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/ml/inference/base.py:984

    # }
    cohort_path_mapping: dict[KeyT, dict[str, list[KeyT]]] = {}
    key_modelid_mapping: dict[KeyT, str] = {}
    seen_keys = set()
    for mp in model_paths:
      keys = mp.keys
      update_path = mp.update_path
      model_id = mp.model_id
      if len(update_path) == 0:
        raise ValueError(f'Invalid model update, path for {keys} is empty')
      for key in keys:
        if key in seen_keys:
          raise ValueError(
              f'Invalid model update: {key} appears in multiple '
              'update lists. A single model update must provide exactly one '
              'updated path per key.')
        seen_keys.add(key)
        if key not in self._key_to_id_map:
          raise ValueError(
              f'Invalid model update: {key} appears in '
              'update, but not in the original configuration.')
        key_modelid_mapping[key] = model_id
        cohort_id = self._key_to_id_map[key]
        if cohort_id not in cohort_path_mapping:
          cohort_path_mapping[cohort_id] = defaultdict(list)
        cohort_path_mapping[cohort_id][update_path].append(key)
    for key in self._key_to_id_map:
      if key not in seen_keys:
        raise ValueError(
            f'Invalid model update: {key} appears in the '
            'original configuration, but not the update.')

    # We now have our new set of cohorts. For each one, update our local model
    # handler configuration and send the results to the ModelManager
    for old_cohort_id, path_key_mapping in cohort_path_mapping.items():
      for updated_path, keys in path_key_mapping.items():
        cohort_id = old_cohort_id

View on GitHub (pinned to 12126d8942)