{"record":{"id":"175e89a8c9a53130","repo":"tensorflow/models","slug":"examples-per-group-self-examples-per-group-is-l","errorCode":null,"errorMessage":"examples_per_group={self._examples_per_group} is larger than the number of examples available in the local (per-device) batch_size={batch_size}. Either decrease examples_per_group or increase the batch_size.","messagePattern":"examples_per_group=(.+?) is larger than the number of examples available in the local \\(per-device\\) batch_size=(.+?)\\. Either decrease examples_per_group or increase the batch_size\\.","errorType":"exception","errorClass":"ValueError","httpStatus":null,"severity":"error","filePath":"official/nlp/modeling/layers/moe.py","lineNumber":588,"sourceCode":"        <float>[batch_size, seq_length, hidden_dim].\n      training: Only apply dropout and jitter noise during training. If not\n        provided taken from tf_keras.backend.\n\n    Returns:\n      Transformed inputs with same shape as inputs:\n        <float>[batch_size, seq_length, hidden_dim].\n\n    Raises:\n      ValueError if we cannot find a group_size satisfying given requirements.\n    \"\"\"\n    if training is None:\n      training = tf_keras.backend.learning_phase()\n\n    # inputs shape [batch_size, seq_length, hidden_dim]\n    batch_size, seq_length, hidden_dim = inputs.shape\n    if batch_size is not None:\n      if self._examples_per_group > batch_size:\n        raise ValueError(\n            f\"examples_per_group={self._examples_per_group} is larger than the \"\n            \"number of examples available in the local (per-device) batch_size=\"\n            f\"{batch_size}. Either decrease examples_per_group or increase the \"\n            \"batch_size.\")\n    tokens_per_group = int(seq_length * self._examples_per_group)\n\n    if training:\n      capacity_factor = self._train_capacity_factor\n    else:\n      capacity_factor = self._eval_capacity_factor\n    # Each group will send expert_capacity tokens to each expert.\n    expert_capacity = int(\n        round(capacity_factor * tokens_per_group / self.num_experts))\n\n    # Reshape batch and sequence/token dimensions for expert routing.\n    x = tf.reshape(inputs, (-1, tokens_per_group, hidden_dim))\n\n    x = self._mask_and_dispatch_to_experts(x, expert_capacity, training)","sourceCodeStart":570,"sourceCodeEnd":606,"githubUrl":"https://github.com/tensorflow/models/blob/e006f5f0d534913e49c1f1dae87364039fa607e2/official/nlp/modeling/layers/moe.py#L570-L606","documentation":"Error \"examples_per_group={self._examples_per_group} is larger than the number of examples available in the local (per-device) batch_size={batch_size}. Either decrease examples_per_group or increase the batch_size.\" thrown in tensorflow/models.","triggerScenarios":"Thrown at official/nlp/modeling/layers/moe.py:588 when the library encounters an invalid state.","commonSituations":"See trigger scenarios.","solutions":[],"exampleFix":null,"handlingStrategy":null,"validationCode":null,"typeGuard":null,"tryCatchPattern":null,"preventionTips":[],"tags":[],"backgroundTag":null,"analyzedSha":"e006f5f0d534913e49c1f1dae87364039fa607e2","analyzedAt":"2026-08-24T14:09:15.576Z","schemaVersion":2},"datasetVersion":"2026-08-24T17:17:21.512Z"}