TensorFlow Transform is a library for preprocessing data with TensorFlow.
tf.Transform is useful for data that requires a full-pass, such as:
- Normalize an input value by mean and standard deviation.
- Convert strings to integers by generating a vocabulary over all input values.
- Convert floats to integers by assigning them to buckets based on the observed data distribution.
TensorFlow has built-in support for manipulations on a single example or a batch
tf.Transform extends these capabilities to support full-passes
over the example data.
The output of
tf.Transform is exported as a
TensorFlow graph to use for training and
serving. Using the same graph for both training and serving can prevent skew
since the same transformations are applied in both stages.
For an introduction to
tf.Transform, see the
tf.Transform section of the
TFX Dev Summit talk on TFX
PyPI package is the
recommended way to install
pip install tensorflow-transform
Build TFT from source
To build from source follow the following steps: Create a virtual environment by running the commands
python3 -m venv <virtualenv_name> source <virtualenv_name>/bin/activate pip3 install setuptools wheel git clone https://github.com/tensorflow/transform.git cd transform python3 setup.py bdist_wheel
This will build the TFT wheel in the dist directory. To install the wheel from dist directory run the commands
cd dist pip3 install tensorflow_transform-<version>-py3-none-any.whl
TFT also hosts nightly packages at https://pypi-nightly.tensorflow.org on Google Cloud. To install the latest nightly package, please use the following command:
pip install -i https://pypi-nightly.tensorflow.org/simple tensorflow-transform
This will install the nightly packages for the major dependencies of TFT such as TensorFlow Metadata (TFMD), TFX Basic Shared Libraries (TFX-BSL).
TensorFlow is required.
Apache Beam is required; it's the way that efficient distributed computation is supported. By default, Apache Beam runs in local mode but can also run in distributed mode using Google Cloud Dataflow and other Apache Beam runners.
Apache Arrow is also required. TFT uses Arrow to represent data internally in order to make use of vectorized numpy functions.
The following table is the
tf.Transform package versions that are
compatible with each other. This is determined by our testing framework, but
other untested combinations may also work.
|GitHub master||2.28.0||2.0.0||nightly (1.x/2.x)||0.28.0||0.28.1|
|0.28.0||2.28.0||2.0.0||1.15 / 2.4||0.28.0||0.28.1|
|0.27.0||2.27.0||2.0.0||1.15 / 2.4||0.27.0||0.27.0|
|0.26.0||2.25.0||0.17.0||1.15 / 2.3||0.26.0||0.26.0|
|0.25.0||2.25.0||0.17.0||1.15 / 2.3||0.25.0||0.25.0|
|0.24.1||2.24.0||0.17.0||1.15 / 2.3||0.24.0||0.24.1|
|0.24.0||2.23.0||0.17.0||1.15 / 2.3||0.24.0||0.24.0|
|0.23.0||2.23.0||0.17.0||1.15 / 2.3||0.23.0||0.23.0|
|0.22.0||2.20.0||0.16.0||1.15 / 2.2||0.22.0||0.22.0|
|0.21.2||2.17.0||0.15.0||1.15 / 2.1||0.21.0||0.21.3|
|0.21.0||2.17.0||0.15.0||1.15 / 2.1||0.21.0||0.21.0|
|0.15.0||2.16.0||0.14.0||1.15 / 2.0||0.15.0||0.15.0|