aboutsummaryrefslogtreecommitdiffhomepage
path: root/google
Commit message (Collapse)AuthorAge
* Updated protobuf submodule to fb714b3 to bring in updates to grpc supportGravatar Manjunath Kudlur2016-02-25
|
* Update protobuf submodule to bring in gRPC supportGravatar Manjunath Kudlur2016-02-23
|
* Enable deserializing protos > 64MB in length.Gravatar Manjunath Kudlur2016-02-18
| | | | | Update protobuf commit Change: 114990608
* Clean up saturate_cast, test, and move to tf.saturate_castGravatar Geoffrey Irving2016-02-11
| | | | Change: 114470777
* Fixed several saver_test errors in open source and with python3.Gravatar Sherry Moore2016-02-11
| | | | | Fixes #1035 Change: 114439527
* Save the graph_def corresponding to the current default graph insteadGravatar A. Unique TensorFlower2016-02-11
| | | | | of a cached version. Change: 114388651
* Run gendocsGravatar Geoffrey Irving2016-02-10
| | | | Change: 114328737
* Remove tf.tensor_util and change tf.unsupported to tf.contrib.utilGravatar Geoffrey Irving2016-02-08
| | | | | | | | tf.tensor_util was already not public (not documented), now it is not part of tf.__all__. The one function that was in tf.unsupported is now tf.contrib.util.constant_value, and there is also tf.contrib.util.make_tensor_proto. Change: 114118183
* Enables `tf.image.resize_images()` to accept tensors for height, width.Gravatar Derek Murray2016-02-05
| | | | | Fixes #1001. Change: 113956896
* Remove the batch_select_op_test I accidentally introduced. This test class wasGravatar Eugene Brevdo2016-02-04
| | | | | moved inside of cwise_ops_test; and this code is outdated. Change: 113857436
* Upgraded to the latest version of Eigen that brings misc performance andGravatar Benoit Steiner2016-02-04
| | | | | stability improvements Change: 113791782
* tf.nn.rnn(..., sequence_length) now conditionally selects states and output.Gravatar Eugene Brevdo2016-02-03
| | | | | | | | | | | | | | | | | | | | | | | | | What this means is that when passing sequence_length to either rnn() or bidirectional_rnn(), the following hold for any given row r of output and state: if time < sequence_lengths(r): output(r) is calculated state(r) is calculated if time >= sequence_lengths(r): output(r) is zero state(r) is carried over from sequence_lengths(r) - 1. What this means when minibatching with sequence_length is that: * When t < min(sequence_length), all the outputs are calculated * When t >= max(sequence_length), all the outputs are zero (and gradients to these outputs will not be propagated) * When min(sequence_length) < t < max(sequence_length), the outputs are a mix of zeros and calculated values (and gradients are conditionally propagated to only those entries which are still calculating) * The final state's rows correspond to: state[r] = state calculated at row r at t = sequence_lengths(r) - 1 and gradients are properly propagated to start updating weights accordingly for each row. Change: 113752785
* Add assert_same_float_dtype.Gravatar A. Unique TensorFlower2016-02-02
| | | | Change: 113646664
* Get rid of mysterious comment previously needed for syncing withGravatar Josh Levenberg2016-02-01
| | | | | the Google-internal version. Change: 113552310
* Add new matrix_solve_ls op for solving linear least-squares problems.Gravatar A. Unique TensorFlower2016-01-26
| | | | Change: 113064195
* Remove TensorShape::ShortDebugString in favor of TensorShape::DebugStringGravatar Geoffrey Irving2016-01-25
| | | | | The two functions already have the same behavior. Change: 112959229
* TensorFlow: Update protobuf submodule to ↵Gravatar Vijay Vasudevan2015-12-10
| | | | 014b9012d32c6409bf2f4f13ff4608a2cbc7cb9a
* TensorFlow: update google/protobuf to newest commit thatGravatar Vijay Vasudevan2015-12-04
| | | | supports python3
* Updated WORKSPACE to point to google/protobuf/gmock.BUILDGravatar Manjunath Kudlur2015-12-04
|
* TensorFlow: update google/protobuf version toGravatar Vijay Vasudevan2015-12-01
| | | | 956a770adc772184e5911f964b866028465d1727 for bazel 0.1.2
* TensorFlow: Initial commit of TensorFlow library.Gravatar Manjunath Kudlur2015-11-06
TensorFlow is an open source software library for numerical computation using data flow graphs. Base CL: 107276108
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-03 22:39:40 +0000