chrisroat
commented
4 years ago FYI. The change to numpy_function was the result of the following warning. Of the two solutions, the switch I made should be equivalent. I believe it just won't allow extra functionality around eager evaluation.
..WARNING:tensorflow:From /Users/chrisroat/flowdec/python/flowdec/tf_ops.py:31: py_func (from tensorflow.python.ops.script_ops) is deprecated and will be removed in a future version.
Instructions for updating:
tf.py_func is deprecated in TF V2. Instead, there are two
options available in V2.
- tf.py_function takes a python function which manipulates tf eager
tensors instead of numpy arrays. It's easy to convert a tf eager tensor to
an ndarray (just call tensor.numpy()) but having access to eager tensors
means `tf.py_function`s can use accelerators such as GPUs as well as
being differentiable using a gradient tape.
- tf.numpy_function maintains the semantics of the deprecated tf.py_func
(it is not differentiable, and manipulates numpy arrays). It drops the
stateful argument making all functions stateful.
eric-czech
Almost all changes are renames. Some of the renames for TF use V1 compatibility API, which means eager evaluation will not work (I believe).
This will break those using older TF, scipy, skimage versions.
Tested in TF 1.14, 1.15, and 2.0, with and without GPU: python/flowdec/*_test.py all pass python/tests/smoketests.sh runs