{"record":{"id":"39e1de7dba32bebf","repo":"tensorflow/models","slug":"the-feature-fusion-method-pyramid-fusion-is-not","errorCode":null,"errorMessage":"The feature fusion method `pyramid_fusion` is not supported in QAT.","messagePattern":"The feature fusion method `pyramid_fusion` is not supported in QAT\\.","errorType":"exception","errorClass":"ValueError","httpStatus":null,"severity":"error","filePath":"official/projects/qat/vision/modeling/layers/nn_layers.py","lineNumber":442,"sourceCode":"    backbone endpoints, and the second is decoder endpoints. When inputs are\n    tensors, they are from a single level of feature maps. When inputs are\n    dictionaries, they contain multiple levels of feature maps, where the key\n    is the index of feature map.\n\n    Args:\n      inputs: A tuple of 2 feature map tensors of shape\n        [batch, height_l, width_l, channels] or 2 dictionaries of tensors:\n        - key: A `str` of the level of the multilevel features.\n        - values: A `tf.Tensor` of the feature map tensors, whose shape is\n          [batch, height_l, width_l, channels].\n\n    Returns:\n      segmentation prediction mask: A `tf.Tensor` of the segmentation mask\n        scores predicted from input features.\n    \"\"\"\n    if self._config_dict['feature_fusion'] in (\n        FeatureFusion.PYRAMID_FUSION, FeatureFusion.PANOPTIC_FPN_FUSION):\n      raise ValueError(\n          'The feature fusion method `pyramid_fusion` is not supported in QAT.')\n\n    backbone_output = inputs[0]\n    decoder_output = inputs[1]\n    if self._config_dict['feature_fusion'] in {\n        FeatureFusion.DEEPLABV3PLUS, FeatureFusion.DEEPLABV3PLUS_SUM_TO_MERGE\n    }:\n      # deeplabv3+ feature fusion.\n      x = decoder_output[str(self._config_dict['level'])] if isinstance(\n          decoder_output, dict) else decoder_output\n      y = backbone_output[str(self._config_dict['low_level'])] if isinstance(\n          backbone_output, dict) else backbone_output\n      y = self._dlv3p_norm(self._dlv3p_conv(y))\n      y = self._activation_layer(y)\n      x = self._resizing_layer(x)\n      x = tf.cast(x, dtype=y.dtype)\n      if self._config_dict['feature_fusion'] == FeatureFusion.DEEPLABV3PLUS:\n        x = self._concat_layer([x, y])","sourceCodeStart":424,"sourceCodeEnd":460,"githubUrl":"https://github.com/tensorflow/models/blob/e006f5f0d534913e49c1f1dae87364039fa607e2/official/projects/qat/vision/modeling/layers/nn_layers.py#L424-L460","documentation":"Error \"The feature fusion method `pyramid_fusion` is not supported in QAT.\" thrown in tensorflow/models.","triggerScenarios":"Thrown at official/projects/qat/vision/modeling/layers/nn_layers.py:442 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"}