Skip to content

Latest commit

 

History

History
79 lines (54 loc) · 2.96 KB

known_issues.md

File metadata and controls

79 lines (54 loc) · 2.96 KB
Raw

Known Issues

The underlying engine behind the decision forests algorithms used by TensorFlow Decision Forests have been extensively production-tested. But this API to TensorFlow and Keras is new, and some issues are expected -- we are trying to fix them as quickly as possible.

See also the known issues of Yggdrasil Decision Forests and the migration guide for behavior that is different from other algorithms.

Windows Pip package is not available

TensorFlow Decision Forest is not yet available as a Windows Pip package.

Workarounds:

Incompatibility with old or nightly version of TensorFlow

TensorFlow ABI is not compatible in between releases. Because TF-DF relies on custom TensorFlow C++ ops, each version of TF-DF is tied to a specific version of TensorFlow. The last released version of TF-DF is always tied to the last released version of TensorFlow.

For reasons, the current version of TF-DF might not be compatible with older versions or with the nightly build of TensorFlow.

If using incompatible versions of TF and TF-DF, you will see cryptic errors such as:

tensorflow_decision_forests/tensorflow/ops/training/training.so: undefined symbol: _ZN10tensorflow11GetNodeAttrERKNS_9AttrSliceEN4absl14lts_2020_09_2311string_viewEPSs

Workarounds:

  • Use the version of TF-DF that is compatible with your version of TensorFlow.

Compatibility table

The following table shows the compatibility between tensorflow_decision_forests and its dependencies:

tensorflow_decision_forests tensorflow
0.1.9 2.6
0.1.1 - 0.1.8 2.5
0.1.0 2.4

No support for TF distribution strategies.

TF-DF does not yet support distribution strategies or datasets that do not fit in memory. This is because the classical decision forest training algorithms already implemented require the entire dataset to be available in memory.

Workaround

  • Downsample your dataset. A rule of thumb is that TF-DF training uses 4 bytes per input dimension, so a dataset with 100 million examples and 10 numerical/categorical features would be 4 GB in memory.

  • Train a manual ensemble on slices of the dataset, i.e. train N models on N slices of data, and average the predictions.

No support for GPU / TPU.

TF-DF does not support GPU or TPU training. Compiling with AVX instructions, however, may speed up serving.

No support for model_to_estimator

TF-DF does not implement the APIs required to convert a trained/untrained model to the estimator format.