- 11 12月, 2018 32 次提交
-
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224872228
-
由 Anna R 提交于
PiperOrigin-RevId: 224870669
-
由 Eugene Brevdo 提交于
This fixes a bug where Adam beta*_power variables were always created as RefVars even if the optimizer acts on ResourceVars. This broke certain defun + Adam use cases. Also fixed the unit tests, which *always* created ResourceVariables (ever since variables.Variable() constructor became aliased to ResourceVariables). PiperOrigin-RevId: 224869338
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224869113
-
由 Shivani Agrawal 提交于
PiperOrigin-RevId: 224865488
-
由 James Ring 提交于
PiperOrigin-RevId: 224863771
-
由 Scott Zhu 提交于
I believe for historical reason, the activation function for LSTM is hard_sigmoid because it is faster compare to sigmoid. With the new LSTM, the performance issue should be fixed with grappler swapping the backend. PiperOrigin-RevId: 224863406
-
由 Skye Wanderman-Milne 提交于
The new toggle replaces ENABLE_COND_V2, ENABLE_WHILE_V2, and ENABLE_TENSOR_ARRAY_V2. This means that these can't be toggled independently anymore, notably that v1 TensorArrays can only be run with v1 loops, and v2 TensorArrays with v2 loops. This also introduces a corresponding environment variable TF_ENABLE_CONTROL_FLOW_V2. I kept the old env vars as well in case people are using them. They all flip the new single toggle now. In addition, this change removes some while_v2 code for dealing with v1 TensorArrays, since this is no longer a supported configuration. PiperOrigin-RevId: 224862245
-
由 Katherine Wu 提交于
Add attribute to Keras model which generates an exportable tf.function. SaveModel save now looks for this attribute when searching for a function to export. PiperOrigin-RevId: 224861089
-
由 Eugene Zhulenev 提交于
PiperOrigin-RevId: 224856604
-
由 Alexandre Passos 提交于
A variable now has a bit which can be turned on which, when turned on, makes that variable act as copy-on-read instead of copy-on-write. This allows sparse writes to happen concurrently while only holding a shared lock, mimicking the use_locking behavior of ref variables. PiperOrigin-RevId: 224855851
-
由 Eugene Zhulenev 提交于
PiperOrigin-RevId: 224854657
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224854195
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224853946
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224853938
-
由 Pete Warden 提交于
PiperOrigin-RevId: 224852926
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224847522
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224844763
-
由 Frank Chen 提交于
PiperOrigin-RevId: 224843723
-
由 Eugene Zhulenev 提交于
PiperOrigin-RevId: 224842783
-
由 Tom Hennigan 提交于
PiperOrigin-RevId: 224841335
-
由 Tong Shen 提交于
Graph optimization passes might overwrite feed/fetch nodes. PiperOrigin-RevId: 224840744
-
由 Pooya Davoodi 提交于
documentaion --> documentation
-
由 Nupur Garg 提交于
PiperOrigin-RevId: 224833908
-
由 Gaurav Jain 提交于
PiperOrigin-RevId: 224833857
-
由 Peter Hawkins 提交于
Fixes crashes seen in JAX test suite on Mac OS. PiperOrigin-RevId: 224832861
-
由 Pooya Davoodi 提交于
Remove NVIDIA link encouraging users to start from https://www.tensorflow.org/install/gpu for installing TensorRT to reduce the confusion about which TensorRT version to install.
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224830708
-
由 Pooya Davoodi 提交于
Tensorflow -> TensorFlow uring -> using
-
由 Dan Moldovan 提交于
PiperOrigin-RevId: 224828066
-
由 Dan Moldovan 提交于
Strengthen the checks in side_effect_guards a bit. This is still not fully robust, but the converter is about to be deprecated anyway. PiperOrigin-RevId: 224827008
-
由 Dan Moldovan 提交于
PiperOrigin-RevId: 224826621
-
- 10 12月, 2018 6 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224800160
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224792124
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224771346
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224757952
-
由 Zhenyu Tan 提交于
PiperOrigin-RevId: 224750894
-
由 TensorFlower Gardener 提交于
PiperOrigin-RevId: 224725432
-
- 09 12月, 2018 2 次提交
-
-
由 A. Unique TensorFlower 提交于
PiperOrigin-RevId: 224696976
-
由 Rick Chao 提交于
Export tf.train.* hooks or hooks-related classes to tf.estimator.* (exporting to both v1 and v2). Keep the existing only in v1. PiperOrigin-RevId: 224658867
-