Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[fix] CI fail on TF2.4.1 #168

Merged
merged 1 commit into from
Oct 13, 2021
Merged

[fix] CI fail on TF2.4.1 #168

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 21 additions & 4 deletions tensorflow_recommenders_addons/dynamic_embedding/python/ops/tf_patch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,8 @@
from tensorflow.python.training import optimizer
from tensorflow.python.training import slot_creator

_PARTITION_SHAPE = 'partition_shape'


class _DenseDynamicEmbeddingTrainableProcessor(optimizer._OptimizableVariable):
"""Processor for dense DynamicEmbedding."""
Expand Down Expand Up @@ -238,6 +240,13 @@ def device_function(self, op):
return worker_device.to_string()


def _assert_float_dtype(dtype):
dtype = dtypes.as_dtype(dtype)
if not dtype.is_floating:
raise ValueError("Expected floating point type, got %s." % dtype)
return dtype


def _compute_fans_for_keras_init_v1_v2(shape):
""" Making keras VarianceScaling initializers v1 & v2 support dynamic shape.
"""
Expand Down Expand Up @@ -300,14 +309,22 @@ def __call__for_keras_init_v1(self, shape, dtype=None, partition_info=None):
def __call__for_keras_init_v2(self, shape, dtype=None, **kwargs):
""" Making keras VarianceScaling initializers v2 support dynamic shape.
"""
kinit2._validate_kwargs(self.__class__.__name__, kwargs)
dtype = kinit2._assert_float_dtype(kinit2._get_dtype(dtype))
if hasattr(kinit2, "_validate_kwargs"):
kinit2._validate_kwargs(self.__class__.__name__, kwargs)
elif hasattr(self, "_validate_kwargs"):
self._validate_kwargs(kwargs)

if hasattr(kinit2, "_get_dtype"):
dtype = _assert_float_dtype(kinit2._get_dtype(dtype))
else:
dtype = _assert_float_dtype(dtype)

scale = self.scale
fan_in, fan_out = _compute_fans_for_keras_init_v1_v2(shape)
fan_in = math_ops.cast(fan_in, dtype=dtype)
fan_out = math_ops.cast(fan_out, dtype=dtype)
if kinit2._PARTITION_SHAPE in kwargs:
shape = kwargs[kinit2._PARTITION_SHAPE]
if _PARTITION_SHAPE in kwargs:
shape = kwargs[_PARTITION_SHAPE]
if self.mode == 'fan_in':
scale /= math_ops.maximum(1., fan_in)
elif self.mode == 'fan_out':
Expand Down