weight decay optimizers doesn't support tf.half on gpu in eager mode #347
Description
System information
- Have I written custom code (as opposed to using a stock example script provided in TensorFlow): No
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux
- TensorFlow installed from (source or binary): binary
- TensorFlow version (use command below): tf-nightly-gpu-2.0-preview
- TensorFlow Addons installed from (source, PyPi): source
- TensorFlow Addons version: master
- Python version and type (eg. Anaconda Python, Stock Python as in Mac, or homebrew installed Python etc): tensorflow/tensorflow:custom-op-gpu
- Bazel version (if compiling from source): N/A
- GCC/Compiler version (if compiling from source):
- Is GPU used? (yes/no): yes
- GPU model (if used):
Describe the bug
Because tf doesn't register tf.half GPU kernel for ResourceScatterAdd op,
https://github.com/tensorflow/tensorflow/blob/b9d106e7ccd9003f39192a77c326fb4ac3e973dd/tensorflow/core/kernels/resource_variable_ops.cc#L968
all of the testSparseRepeatedIndices with tf.half failed in:
https://github.com/tensorflow/addons/blob/master/tensorflow_addons/optimizers/weight_decay_optimizers_test.py
I find keras still uses test_util.run_deprecated_v1 for Adam test case
https://github.com/tensorflow/tensorflow/blob/560d21b68805ed3247f897b9ac2e1f203a168f8d/tensorflow/python/keras/optimizer_v2/adam_test.py#L113 , so I guess _resource_scatter_add still cannot work well with tf.half on gpu in eager mode.
Describe the expected behavior
Code to reproduce the issue
Other info / logs