hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
7042de9de41261df8f5c2d9cca86ee24efd9cf92 | Merge pull request #728 from ageron/master. | Martin Wicke | wicke@google.com | 1,456,419,800,000 |
e752109efb67c07d1017646182e75268e27ae4fb | Fixes bug in accumulation of total-approximate-duality-gap.
Change: 115528686 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,380,385,000 |
c54b03ad9c6bbb40d6a3adb2ca6a43333cde87e0 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,456,420,337,000 |
d16868d4bd4c7730b9d0b02f97ee9ea46b77f057 | Fixes #110: Corrected neural_translation_word example to run and train a translation model. | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,423,816,000 |
82ecfff7daf984191cb7ee6818314c82caa5e6aa | Fix for constant folding where nodes with no inputs doesn't get constant folded.
Change: 115568214 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,419,868,000 |
5c9f4f8973d0958d13e30be574f9403df43b87d7 | TensorFlow: fix bug in StringPiece::contains which made it always
return true. Add a unittest to catch this type of regression in
the future.
Change: 115573280 | Vijay Vasudevan | vrv@google.com | 1,456,423,054,000 |
63bd3efc5c9a052d605766497726120b73f5cb3c | Made sure that the tracking allocator always counts the allocated sizes.
Made the corresponding unit test more robust.
Change: 115575179 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,456,423,946,000 |
9ccc4b6afe4defa748de89eac90ba1062232bb5a | Avoid some over-inlined routines. Reduces code size of TensorFlow binaries
considerably. Shrinks text size of example_trainer binary by ~1.5%.
Change: 115578002 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,425,426,000 |
d1aed6505a7703ca9f596d8d415257fd238a62c6 | Add contrib/testing.
Change: 115578243 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,425,557,000 |
86e93febaa2f349f2fbf733d6895a69e87ec578e | TensorFlow: Fix scatter_op_test now that StringPiece::contains is fixed.
Change: 115580211 | Vijay Vasudevan | vrv@google.com | 1,456,426,521,000 |
13d7f5203454d328c6cc6cb048cc3682cc53fe9e | TensorFlow: make split_op not use internal header library for callback,
since this breaks the build on GPU.
Change: 115582331 | Vijay Vasudevan | vrv@google.com | 1,456,427,451,000 |
818644c2a946a9d73262ca67b17b98d54856377f | Changed testing::SrcDir() to testing::TensorFlowSourceRoot() and fixed it.
Also fixed some compiler warnings.
Change: 115582482 | Derek Murray | mrry@google.com | 1,456,427,523,000 |
12da3ed3a95719bf7ed8e2be31b2fa911ba76ed0 | Fix predict data feeder to return non-overlapping batches
setup_predict_data_feeder when give an iterator returns
incorrect batches. The incorrect batches take the following
form.
[0, ..., batch_size]
[0, ..., batch_size, batch_size+1]
...
This commit fixes this to return non-overlapping batches of
sizes equal to ba... | Ali Elqursh | elqursh@google.com | 1,456,427,863,000 |
7b47c8b4a362556bdbd604c81f804d76893737e4 | Add native depthwise_convolution op (forward pass).
The current depthwise_conv is very inefficient by calling slice() on each
input channel on input and filters, followed by a conv() on each input channel,
after which is a concat().
Change: 115583330 | Jianmin Chen | goog.jmchen@gmail.com | 1,456,427,945,000 |
356bf7f4669c545486da9b862ef745fc857a1e21 | TensorFlow: add missing header file to posix/test.cc
Change: 115589382 | Vijay Vasudevan | vrv@google.com | 1,456,431,017,000 |
590b8c6c3986f0e494657d85848e5ae0d59d550b | Make estimator.predict more memory efficient
For multi-class classification estimator.predict currently
appends all the predictions and then argmax the resulting
tensor. For a large number of classes this tensor is huge
and may not fit in memory.
This CL changes the behavior to argmax before appending
the results for... | Ali Elqursh | elqursh@google.com | 1,456,431,740,000 |
b97931cb933d546b578a17d3454e9f4eb3b91cad | Updated protobuf submodule to fb714b3 to bring in updates to grpc support | Manjunath Kudlur | keveman@gmail.com | 1,456,432,784,000 |
c54829b909555c0836d35c8ada3fea5abe5e7ce0 | Fix iter.next compatiblity with python3 | Ali Elqursh | elqursh@google.com | 1,456,434,537,000 |
d1245c3c87160760c0fb66f19ddbd7fa48989e81 | Merge pull request #115 from elqursh/optimize-predict
Make estimator.predict memory efficient for large n_classes | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,434,861,000 |
ad3ef4c05b2fab077ce7fb56a4927100fed239e9 | TestReporter is back in. Maybe also fixed the Android build.
Change: 115589642 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,431,167,000 |
eec5477ab6e5f05a7e6a2d3e1f9549bafd7dfa09 | Execute TODO to rename io.* to save_restore_tensor.*. This will
hopefully reduce confusion since io.* is not the implementation of the
".../kernels:io" build target.
Change: 115593814 | Josh Levenberg | josh11b@tensorflow.org | 1,456,433,699,000 |
be64da94541cf5bb63e09c110a321d505122e0fa | Remove endl at the end of VLOG.
Change: 115594986 | Manjunath Kudlur | keveman@gmail.com | 1,456,434,335,000 |
a5f3979004299679b14f45812afd2fc91f27f7d9 | Clarify comments for max_to_keep.
Change: 115598592 | Sherry Moore | sherrym@google.com | 1,456,436,300,000 |
a82f7e6b55091d29896e0767ae7fb06f93447d86 | Make gpu_lib for non-cuda deps that we use in public kernels.
Change: 115598732 | Vijay Vasudevan | vrv@google.com | 1,456,436,364,000 |
97f6b6fb66cd8e52af4750bd183dfc555d08ef4d | TensorFlow: perl command in configure script was not properly replacing
the cuda/cudnn version strings in build_config.bzl
Addresses issue mentioned in #1052
Change: 115599314 | Vijay Vasudevan | vrv@google.com | 1,456,436,646,000 |
9f58b2404aed5dc3ef4cc58a0caff6f37a63b9bb | Merge pull request #114 from elqursh/fix-data-feeder
Fix predict data feeder to return non-overlapping batches | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,456,436,903,000 |
90cf3e2eeaddd480cec8a587b8b20b3b562427ef | Rollback of: Add native depthwise_convolution op (forward pass).
The current depthwise_conv is very inefficient by calling slice() on each
input channel on input and filters, followed by a conv() on each input channel,
after which is a concat().
Change: 115601904 | Vijay Vasudevan | vrv@google.com | 1,456,437,865,000 |
c38bbf42e800f0ff8eb870528053a37f7715134a | Rollback of "TestReporter is back in. Maybe also fixed the Android build."
Test fails.
Change: 115602477 | Vijay Vasudevan | vrv@google.com | 1,456,438,126,000 |
1fd7428866978b2b98431a1bd88be1a7b9e6162d | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,456,438,434,000 |
f3ead2df04a93e97d9e247fa1691f03680ef2911 | Correct handling of argv in test utility.
Change: 115607801 | Derek Murray | mrry@google.com | 1,456,440,947,000 |
03fed366e4288e16e34d7f99ffb5fb842123b40c | TensorFlow: conv_ops uses gpu_device_context, needs to depend on the lib.
Change: 115607974 | Vijay Vasudevan | vrv@google.com | 1,456,441,041,000 |
aeae4825b35eb8cb18b5478a1d8863272b8a3634 | Add symbolic gradient functions for Conv2D and MaxPool
Change: 115608522 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,441,374,000 |
cdd0f2eeef9a11a48433156e41c95a5fd6f4e1ee | Fix compilation error in argv parsing code... whoops.
Change: 115610448 | Derek Murray | mrry@google.com | 1,456,442,385,000 |
e868703dd06193002ad98b4c43217b56013e6db2 | Add example accessing of weights
It wasn't clear how to access weights using
`classifier.get_tensor_value('foo')` syntax. This adds some
examples for the CNN model. They were figured out by
logging the training as though for using TensorBoard, and then
running `strings` on the logfile to look for the right namespace... | Dan Van Boxel | the.dvb@gmail.com | 1,456,243,799,000 |
fd974e09eef3bc5c3d69f6304d4d1083e90a552c | Merge pull request #111 from dvbuntu/custom_weights_example
Add example accessing of weights | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,456,451,030,000 |
8e4858e08915aff45b87f06ad0c719e7d8397c1f | Typo fix and add example to TOC | terrytangyuan | terrytangyuan@Gmail.com | 1,456,451,393,000 |
1fa740f7f638bc2ec94e4cb278f60c34afaab206 | Added recent changes | terrytangyuan | terrytangyuan@Gmail.com | 1,456,451,970,000 |
01a6f5e504d9299395888a786e52c589c16af529 | Multiple layout support for pooling operations.
Change: 115611259 | Xiaoqiang Zheng | zhengxq@google.com | 1,456,442,830,000 |
82fdff99d420ead692b5256d993448da97e4988a | Update TensorBoard dependencies versions.
Remove some unnecessary devDependencies, e.g. accessibility_developer_tools and chai.
Change: 115613184 | Dan Mané | danmane@gmail.com | 1,456,443,902,000 |
b62169037fcbb7e56c73839e6314ad79050694a8 | Split TensorBoard gulpfile into per-task files.
Gulp task changes:
"compile.all" -> "compile"
"tslint-strict" -> "tslint"
"tslint-permissive" -> "tslint.permissive"
gulp watch will run tslint and compile in parallel rather than in sequence (faster)
Change: 115614194 | Dan Mané | danmane@gmail.com | 1,456,444,471,000 |
841656c9fd84902733e2ee13ac4dd9020a5b6361 | Change ReaderOp kernel to be async, by having it own a thread on which it
schedules callbacks to run.
Change: 115615528 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,445,208,000 |
b51ef0cd06e1bfb529b272e55010790ff3ead31f | Rollback of: Add native depthwise_convolution op (forward pass).
The current depthwise_conv is very inefficient by calling slice() on each
input channel on input and filters, followed by a conv() on each input channel,
after which is a concat().
Change: 115617901 | Jianmin Chen | goog.jmchen@gmail.com | 1,456,446,478,000 |
ca69460901cd8f6108c2e6cda6974a8da4efde6b | Update TensorBoard readme with info on exporting data.
Change: 115619523 | Dan Mané | danmane@gmail.com | 1,456,447,449,000 |
e8260f34332a90b3de91e2317909f26bcd1d5936 | Add a step_id field to the OpKernelContext indicating which step
triggered the Op execution. This will allow us to "charge" resource
usage to a specific step.
Change: 115622649 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,449,425,000 |
d8fbe760f02662db7c79969aec15227fc261f8f5 | Make TensorBoard purging of orphaned or out-of-order events optional.
Now it is controlled by a flag.
Default behavior is the same.
Will be quite useful for debugging missing data type issues.
Change: 115623272 | Dan Mané | danmane@gmail.com | 1,456,449,864,000 |
9f7eb520c750303549ca75a63b57924f2a99ef66 | Missing entries due to moved files.
Change: 115626850 | Vijay Vasudevan | vrv@google.com | 1,456,452,414,000 |
657d5fecbfb14c9f86a4a84ad09e78642397dd77 | Rollback of:
Make TensorBoard purging of orphaned or out-of-order events optional.
Now it is controlled by a flag.
Default behavior is the same.
Will be quite useful for debugging missing data type issues.
Change: 115627868 | Vijay Vasudevan | vrv@google.com | 1,456,453,275,000 |
fb2d3884b0e188b5f8bb55ba13608712bc989e6a | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,456,454,684,000 |
7dd05309b379fd49677a5c795d95664895bb16ca | Reduce log output for each step in direct_session, added previously, from LOG(INFO) to VLOG(1).
Change: 115630069 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,455,366,000 |
045e08676489335b0e294541d268d0fa09a2f0c9 | Rollback of "Split TensorBoard gulpfile into per-task files.
Gulp task changes:
"compile.all" -> "compile"
"tslint-strict" -> "tslint"
"tslint-permissive" -> "tslint.permissive"
gulp watch will run tslint and compile in parallel rather than in sequence (faster)"
Change: 115630394 | Vijay Vasudevan | vrv@google.com | 1,456,455,703,000 |
d27da251bcc4bab7da2f5aecc509b146f9fa1692 | TensorFlow: fix GPU breakage (again)
Change: 115631502 | Vijay Vasudevan | vrv@google.com | 1,456,456,737,000 |
5a30c8f07ebc4817c16399416891bfc95304fda7 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,456,456,857,000 |
00986d48bb646daab659503ad3a713919865f32d | Initial version of the open-source distributed TensorFlow runtime.
This includes a gRPC server (grpc_tensorflow_server) that can serve as both
the master of a distributed TensorFlow computation, and an individual worker
in the computation. The GrpcSession class is included to allow client programs
(including Python cl... | Derek Murray | mrry@google.com | 1,456,459,809,000 |
b88971051fbc49fa1e0b91ec1b0b60defa11697e | Merge commit for internal changes | Derek Murray | mrry@google.com | 1,456,492,115,000 |
c487eb53b925d23c6df005e3fea29b8ea82486ae | Fixing nightly artifact version in README.md | Shanqing Cai | cais@google.com | 1,456,515,176,000 |
5c5b29adda87f6fbedd1e629d516af317358d556 | mnist example now uses sparse_softmax_cross_entropy_with_logits.
Change: 115636069 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,461,874,000 |
5d01083d62a24b9a525f7869595385608cb1dbdc | Add an argument to control memory swap for while loop gradient.
Change: 115637967 | Yuan Yu | yuanbyu@google.com | 1,456,463,958,000 |
4f59a43219c25d99b7e62b3964afed3d463f65f8 | Move core/common_runtime/gpu_device_context.h into
core:gpu_lib, out of the non-gpu core:core_cpu* targets.
Change: 115641392 | Josh Levenberg | josh11b@tensorflow.org | 1,456,468,441,000 |
e0d5b85a1697f986f21559bbbeea3c42d4b66a35 | Use gfile consistently.
Change: 115663554 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,495,951,000 |
dff7bb7c736787b05d5cd24aaf64cb787a68ca53 | Adds a number of loss functions and reduction utilities to calculate quantities such as L1 and L2 loss.
Change: 115671496 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,503,824,000 |
fdfbd3af7a25410a999cbd6a0733e36ddd1ddb14 | Final fix to TestReporter (hopefully).
Change: 115675044 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,506,375,000 |
8509f88ade3fc65f4cc530780d06100f0b1ea108 | Add CTC (Connectionist Temporal Classification) Ops to TF contrib.
This includes:
* ctc_loss
* ctc_greedy_decoder
* ctc_beam_search_decoder
Change: 115683564 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,510,926,000 |
c36725ad792737ed74d2c02743d70f704730858e | SDCA implementation changes in preparation for larger data-set scaling.
Change: 115686672 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,512,648,000 |
96118731cb793f0d51adfb7a4df058ee3c5fa00a | Adds RunWithOpts() that takes a per-step RunOptions and RunOutputs to
C++ Session API. Use it for optionally turning tracing on for a step and
returning profiling info collected via StepStats.
For DirectSession only.
Example usage:
RunOptions run_options;
run_options.set_trace_level(RunOptions::FULL_TRAC... | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,516,168,000 |
f417134a5a41a9fc4ab0161c9215633f3c4c0156 | Copy protobuf header files and Eigen to the pip package.
Clarify the documentation on how to build using bazel.
Fixes #1270.
Change: 115701645 | Manjunath Kudlur | keveman@gmail.com | 1,456,521,406,000 |
25a5fbcf642c5b65ce1302557c3e77db377a62dd | Allow list of initializer nodes to run before freezing the graph.
Change: 115703618 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,522,496,000 |
8d7f58d44d9e3d2d82066cc5c286d6d433079621 | Make sparse index error messages more informative
They now say what the index is and give the required bounds for out of bounds
errors. Example (slightly paraphrased):
Index 0 is out of bounds, foolish user.
vs.
indices[0] = [0,128] is out of bounds: need 0 <= index < [64,10]
Change: 115705692 | Geoffrey Irving | geoffreyi@google.com | 1,456,523,668,000 |
6390a06d803fab48536d13a05ffb68d24b4cc991 | Fix small dtype inference error in ctc loss op test.
Change: 115706184 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,523,950,000 |
8bc340fd232c914bee956a4c8ff0d83a52757fd6 | Merge pull request #1305 from caisq/ver-update-2. | Vijay Vasudevan | vrv@google.com | 1,456,525,528,000 |
31fd5b0c998ad4c79a90bab1555303e57e1b0f08 | Corrected unbalanced parentheses in docs. | Konrad Magnusson | konradma@kth.se | 1,456,499,799,000 |
77c2042e77a11ee442ecc7e369cd91d91e4a98c3 | Merge pull request #1302 from KonradMagnusson/master. | Vijay Vasudevan | vrv@google.com | 1,456,525,590,000 |
d99245f1ccd6be547702c3e4d4e8f8e54597c569 | Fix examples given new tensorflow rnn interface. Update bidirectional rnn to return final state. | Danny Goodman | goodman.danny@gmail.com | 1,456,546,889,000 |
a723b61abd25ae3045f862015670d0bf00959e40 | fix spacing | Danny Goodman | goodman.danny@gmail.com | 1,456,548,741,000 |
8469188fb1753b72b29af2a8f0439c207bb97fc5 | update travis ci version of tensorflow | Danny Goodman | goodman.danny@gmail.com | 1,456,550,215,000 |
6becc9eabeb7ff18a7f055374cbfe5d7bdf5855c | Assign graph to cpu in example/udacity/assignment5.
This was required so the code will execute on tensorflow
installations with gpu support. SparseApplyAdagrad op has
no gpu implementation at this time and causes exception to be
thrown. | J Yegerlehner | jyegerlehner@yahoo.com | 1,456,649,079,000 |
1d4f00da15a886916cd7a62ddf119b0b460c850c | Merge pull request #1315 from jyegerlehner/udacity-5-gpu-fix. | Vincent Vanhoucke | vanhoucke@google.com | 1,456,675,184,000 |
20c562f56120c377a1911a0b1eb7327753f771f0 | Added some robustness to interrupting jobs while save is been run (before would just leave with an empty file, now there is a backup) | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,705,261,000 |
b185f114c715e7bc0c54d599f636a28e3e78497a | Merge branch 'master' of github.com:google/skflow | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,705,798,000 |
e576ca6eb5ef5aaff670785530b842feb3d81316 | Fixes #118: Trim in CategoricalVocabulary remaps ids to 1..n to make sure that max number can be used in categorical variables. | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,706,812,000 |
7a7bbcc8b47f54384a43494ac78083de9e3915e4 | Merge pull request #119 from dgboy2000/master
Fix examples given new tensorflow rnn interface. Update bidirectional… | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,707,109,000 |
abdfb10f2887210d18ff3b33596085f104029d7d | Up the required version of Tensorflow - due to recent RNN changes, skflow will now only support 0.7+ | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,708,821,000 |
d28ca29e52f7635f24470d696c4b61ec26661fa0 | Ref #118: Additional fixes to neural_translation_word example. | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,708,974,000 |
7b11cedae8d9547fe5401538bb22b973e4361abc | Fixing changes in tests due to reordering after trimming | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,709,179,000 |
eeeb65945f171e5b5d7397b8a3ea2112402234fe | Make vocabulary sorting by alphabet first to make it a persistent sort between python versions for the same frequency. | Illia Polosukhin | ilblackdragon@gmail.com | 1,456,712,123,000 |
2954fcea8e5abacc65d06b732c94cc848465e443 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,456,761,757,000 |
6f74cd15a60821211444acfedc8e5e68782a6fcf | Refactor DirectoryWatcher to take in a path provider as input.
This will allow us to watchings things that aren't local files on disk, such as
GCS directories.
Change: 115708990 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,525,462,000 |
81f3078debd52f9bdba4cf2723c39ebc60fbc21d | Delete deprecated core:kernel_lib target now that it is no
longer referenced.
Change: 115711086 | Josh Levenberg | josh11b@tensorflow.org | 1,456,526,661,000 |
1df80e82acd1209f6cc95e68c9eedc59953d4a95 | Minor: default-construct RunOptions and RunOutputs in vanilla Run().
Change: 115714930 | Zongheng Yang | zongheng.y@gmail.com | 1,456,528,908,000 |
1827f66ab8329694d2f7831d83d12c10a8d7b804 | Bugfix to dynamic_rnn: restore some shape information when possible during loop.
Change: 115717351 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,530,272,000 |
e9c1ccd421dd4e669dc88fd8cb4e1c49e6ea1933 | TensorFlow: make EqualGraphDef ignore internal attributes (those that
starts with "_", which are not part of the public interface of ops and
can change.
Change: 115719866 | Vijay Vasudevan | vrv@google.com | 1,456,531,752,000 |
7552e9e56156cce94459c55915657e641f07db80 | Add build ruiles to CTC
Change: 115720167 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,531,953,000 |
ab48dbd4ac2095548a5bc8505e08e751d409727f | Fixing operator order in LRN docs to match code.
The implementation adds the bias to a temporary, that temporary is what is then the base with exponent beta. The implementation also agrees with the equation in Section 3.3 of the referenced Krizhevsky et. al. paper.
Change: 115721267 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,532,629,000 |
52c73af0b029f69b5435fec4c8e6ff68f6bda4fd | TensorFlow: add 'make_ndarray' to contrib/util
Change: 115722090 | Vijay Vasudevan | vrv@google.com | 1,456,533,195,000 |
06b0813ddde0451e56d2fe8fc55711d30cb1b45f | Initial API for tf.contrib.distributions.
Change: 115725802 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,535,942,000 |
c04d14e09e1f5e09259b37a133ac31c2217f94cd | Adding an op to merge ids and values as output by parse_example into a single SparseTensor.
Change: 115729905 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,540,006,000 |
fdabb9a0eeaa6885d68ff314a12edda5679eca91 | Bugfix to the Any protobuf
Change: 115740568 | Eugene Brevdo | ebrevdo@gmail.com | 1,456,557,278,000 |
a0c9e2d6d5115aba04a650281b10d47e31873671 | Fix futures test
Change: 115766190 | Martin Wicke | wicke@google.com | 1,456,616,349,000 |
c2a7ea886e4ef49a870481007c209d32f317cd14 | Compute non-approximate duality gap based on the previous iteration
Change: 115796400 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,685,170,000 |
db828ab20399697bb97c218ca6c435ee59b2a029 | SDCA now accepts example_ids.
Change: 115850355 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,761,757,000 |
977ea14361f7bceeeda2046f3073ea888806be02 | Added more streamlined interfaces for converting rendezvous ids
to/from strings and used these in the rendezvous code. Improves
performance for ptb_word_lm slightly (saves several allocations and an
sscanf per CPU <-> GPU transfer).
Change: 115852277 | A. Unique TensorFlower | nobody@tensorflow.org | 1,456,762,678,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.