Chris Schmitz

Infer_Schema in FDW Utils: Can't handle features with no shape info

1

Infer_schema always creates FixedLenFeatures and populates their shape with the shape from the dataset's element_spec. FixedLenFeatures require a shape, but in the element_spec that shape is often None, I assume...

TF Dataset to TF Examples List in FDW Utils: Can't handle sparse tensors / empty values

1

Datasets where a column relevant to FDW (specifically, a column to slice on) contains empty values throw errors in one of two places: - If the dataset feature map contains...

The current implementation of TF Dataset to TF Examples List uses a lot of RAM and runs very slowly on large datasets. In [this colab](https://colab.research.google.com/drive/12SNg529v0WiMBY8LktuSIvurFV5RUYsC#scrollTo=dpD3CjebGGHA&line=1&uniqifier=1), the CelebA dataset, as processed...

The tf_dataset_to_tf_examples_list function in fdw utils [here](https://github.com/tensorflow/model-remediation/blob/e36bc4e497cf731d66a27b8266b9b47cf8618648/tensorflow_model_remediation/experimental/fair_data_reweighting/utils.py#L157) can only handle datasets where each element is just a neat single-layer dict of format `{feature_name: tf.Tensor}`. The easiest way to generate one...