hash
stringlengths
40
40
msg
stringlengths
1
131k
author
stringlengths
1
33
email
stringlengths
0
57
date
int64
1,447B
1,698B
98866b84ac21828b1964e7bdb1d0396091ac9438
Adding new category LOCAL_INITIALIZERS to GraphKeys. This category collects initializers that should run for every new instance. Change: 115868141
A. Unique TensorFlower
nobody@tensorflow.org
1,456,770,357,000
a886c9e74a00038d12866c60b3c2e5efcc4ee8e4
TensorFlow: fix undeclared inclusion for conv ops. Fixes #1334 Change: 115873803
Vijay Vasudevan
vrv@google.com
1,456,772,475,000
cfcf3d6066a41c61d6cf73b328af5099034cbe86
Fix OSS build breakage due to missing std namespace on vector. Change: 115879158
A. Unique TensorFlower
nobody@tensorflow.org
1,456,774,625,000
9f7fc90713cefb2c45467b0b0e487e2003e9001d
Generalize tf.test.compute_gradients to handle complex inputs and outputs If inputs or outputs are complex, the Jacobian will still be real but the corresponding Jacobian dimension(s) will be twice as large. This is required even if both input and output is complex since TensorFlow graphs are not necessarily holomorp...
Geoffrey Irving
geoffreyi@google.com
1,456,777,134,000
d12c2dab70b65d616fc7649f932ff8a585b9aca8
Fix mixing type sort in categorical vocabulray
Illia Polosukhin
ilblackdragon@gmail.com
1,456,787,007,000
52e128fbd18245da274471627ee7e640d5cd3098
Improved the performance of narrow reductions on CUDA Change: 115889721
Benoit Steiner
benoit.steiner.goog@gmail.com
1,456,779,415,000
384480d0fa3cfa3eaed7e6f685d10c20bcf14a6f
Added a benchmark for the performance of dyanmic rnn with memory swapping. For the large model, the overhead is roughly 35%. More improvements over this base-line implementation are coming. But if you have long sequence models that are currently running out of memory, I encourage you to try this out. Calculation: Dyn...
Yuan Yu
yuanbyu@google.com
1,456,788,858,000
de50fc8cc523b3403ba54a29174b05a771971b99
Make TensorBoard purging of orphaned events optional. It's now controlled by a flag. Update README with some information. This is a resubmission of 115623272, with tests passing :P Change: 115913395
Dan Mané
danmane@gmail.com
1,456,789,406,000
d4b17f62fa09a10869d0f5d70e5afde833b23f0f
Expose tf.contrib.util.stripped_op_list_for_graph C++ and Python use two different op registries, and in rare cases they can actually be different. Thus, we need both functions available. Also fix both Python and C++ to handle arbitrarily nested functions. Change: 115915068
Geoffrey Irving
geoffreyi@google.com
1,456,790,127,000
c1d385d01924cf02e81adabbdec575c100b14f59
Expose tf.contrib.util.stripped_op_list_for_graph C++ and Python use two different op registries, and in rare cases they can actually be different. Thus, we need both functions available. Also fix both Python and C++ to handle arbitrarily nested functions. Change: 115918836
Geoffrey Irving
geoffreyi@google.com
1,456,791,656,000
66feae1e093708a4e124f61d72af0bf7c5433eb9
fix gen_docs_test new methods xx in variable.py have to be mentioned with @@xxx in state_ops.py Change: 115926888
A. Unique TensorFlower
nobody@tensorflow.org
1,456,795,722,000
0c69be4e12fe5355cfdba931780a83d666e7abea
Lazily initialize the static coeffcients table once, on first use, rather than doing it every time ResizeBicubicOp<...>::Compute is invoked. Also refactored the way this is handled so that there is at most one coefficients table in the process, rather than one per TensorFlow numeric type. Saves about 60K bytes / bina...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,795,868,000
7f12947e4f31cdf9a0cca291a653980fa204d686
Don't provide a padding op for strings: the code never did what could be reasonably expected (i.e. pad with spaces), and doesn't compile anymore. Change: 115936929
Benoit Steiner
benoit.steiner.goog@gmail.com
1,456,802,018,000
68bc6ef4d9cc246f3d2d73036a4c9cd47f87198d
Now using tf.merge_sparse to create SparseTensors. Change: 115944708
A. Unique TensorFlower
nobody@tensorflow.org
1,456,808,045,000
e928a43f79163bdbbe3cca6c240ea4630f75b1b0
Merge commit for internal changes
Vijay Vasudevan
vrv@google.com
1,456,816,188,000
e657a1b115c56257cb194447c34852e3c6ff2688
DeltaWeights and other refactoring/simplification. Change: 115957374
A. Unique TensorFlower
nobody@tensorflow.org
1,456,818,924,000
b626859fbfda39141d0adb50558bf2aa5fb60cd9
Merge commit for internal changes
Manjunath Kudlur
keveman@gmail.com
1,456,844,850,000
48b3b14cf0a8f47579801eb58df3737f4548fce0
fix of the rnn_seq2seq in order to be compatible with tensorflow 0.7
Zikes, Jan
jan.zikes@merck.com
1,456,848,984,000
67d3fb03efdfa37e2b7013e2802d1b737f8fbe85
Merge pull request #123 from ziky90/master fix of the rnn_seq2seq in order to be compatible with tensorflow 0.7
Yuan (Terry) Tang
terrytangyuan@users.noreply.github.com
1,456,851,945,000
23753679279336f425484185a03fc3747fc79361
Update retrain.py to be compatible with Python 3 It's PROBABLY backwards compatible with Python 2 but I haven't tested it Fixes https://github.com/tensorflow/tensorflow/issues/1274
Daniel Golden
dan@arterys.com
1,456,356,852,000
2ce463c1a750ca37b93327f05a37914effa96b6b
Merge pull request #1290 from dgolden1/retrain-py3.
Vijay Vasudevan
vrv@google.com
1,456,874,446,000
d89b25bf6455b6b4a5da7d1d489d6d24d59cdd45
Fix PCRE link in get started docs
Michael Peteuil
michael.peteuil@gmail.com
1,456,688,710,000
eddddc684a0a17345636801f396a56f9af983c00
Merge pull request #1327 from mpeteuil/fix-md-pcre-link.
Vijay Vasudevan
vrv@google.com
1,456,874,501,000
d39a592c6da3b0f3a3fadd64ab98fc0712b830b6
Fix typo in README
Isaac Hodes
isaachodes@gmail.com
1,456,813,527,000
aaf471a5f40923134c5e0aaffd17f3fe6b2ed1cb
Merge pull request #1339 from ihodes/patch-1.
Vijay Vasudevan
vrv@google.com
1,456,874,955,000
75b11620d576bd38b3a2ab84ae3b777a9ac35937
Fix typo
Johnny Lim
izeye@naver.com
1,456,803,944,000
acb921ffbf2421798b69da7f1b4f885ec19766e4
Merge pull request #1337 from izeye/patch-5.
Vijay Vasudevan
vrv@google.com
1,456,875,015,000
c6c063bce1c617371c4ff6254b90a294ad3d360c
Fixing a still-broken link to artifact in README.md
Shanqing Cai
cais@google.com
1,456,763,165,000
b67f8d3966c89134aed26a0821cfb80c1cfe65f2
Merge pull request #1332 from caisq/ver-update-2.
Vijay Vasudevan
vrv@google.com
1,456,875,090,000
65ac23d1e0209c699e4454e749ab85443207fb71
Add _python_build to .gitignore The directions for a development build say to make it, so .gitignore should ignore it.
Geoffrey Irving
geoffreyi@google.com
1,456,610,190,000
c35de99b9351414a27f0802acef61828a8c5d65b
Merge pull request #1320 from girving/ignore.
Vijay Vasudevan
vrv@google.com
1,456,875,150,000
999b64fc9f6ea10c15f8a5301c0a8c19a110c8a8
Removed a broken link in index.md A link referred to CIFAR-10 samples. However, this was a broken link. So, I removed the link and the references to this link.
Pranav Sailesh Mani
pranav656@gmail.com
1,456,582,317,000
6001868e01c72dd6c54b7ffd5c3aaf66313b6f8d
Merge pull request #1316 from delftswa2016/removed-broken-link.
Vijay Vasudevan
vrv@google.com
1,456,875,204,000
20b444f30b4db29edde1e8f582b489ac8d1abd2d
Fix typo 'overlaps' -> 'overlap'
Aggelos Avgerinos
evaggelos.avgerinos@gmail.com
1,456,534,698,000
ceeef1c4da658e5b0891c7071915ca337b4bf581
Merge pull request #1312 from eavgerinos/typo-fix.
Vijay Vasudevan
vrv@google.com
1,456,876,997,000
75895b03177ed68206d1ce61d4bf746d551bd914
did grammatical clean up
ninotoshi
ninotoshi@users.noreply.github.com
1,456,486,125,000
83fd02de324c8b3f65c8906b221cff0097c961d5
Merge pull request #1296 from ninotoshi/master.
Vijay Vasudevan
vrv@google.com
1,456,877,052,000
77d4e89afa8b1940ff67c44702c048b2d6b747f9
Fix word2vec_basic.py example to be python 3 compatible
timsl
tiol@kth.se
1,456,779,263,000
342377a8a7fa56501d716b7cef5384f7b3483b17
Merge pull request #1335 from timsl/py3-example.
Vijay Vasudevan
vrv@google.com
1,456,877,660,000
77887d9f79d0573ceb6254cd322fdbe79f106978
Fixes documentation in distributed runtime README. The flag `--task_index` should be `--task_id`, as noted in the discussion of Issue #1310. Also deletes an unused file. Change: 115996182
Derek Murray
mrry@google.com
1,456,846,085,000
fc2abdc474cb3860c95fe6a3affd569077954f60
Added support for returning a tuple in tf.py_func. Change: 116009049
A. Unique TensorFlower
nobody@tensorflow.org
1,456,853,112,000
1fd3da09401168d2c4ed1ff68494644fcf03ba92
Delete RequireDefaultOps() now that all references have been deleted. Change: 116009282
Josh Levenberg
josh11b@tensorflow.org
1,456,853,207,000
19b7fd80780d02372f76076bc8eb40d55a89a301
Improves the syntax to define a tf function w/ a list of arguments. Previously, @Defun(x=tf.float32, y=tf.float32) def Foo(x, y): return x + y Now, @Defun(tf.float32, tf.float32) def Foo(x, y): return x + y Supports varargs. E.g., @Defun([tf.float32]*10) def Foo(*args): sum = args[0] for x in args[1:]: s...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,854,191,000
61c3d25c07f38f130e185a6731a86341012f6bc8
Avoid parallelism in the optimizer unit-test, as each step is stochastic, the optimizer may produce different results (and multi-threading aggravates it). Change: 116028506
A. Unique TensorFlower
nobody@tensorflow.org
1,456,860,950,000
43b89c901621d8db4a534e32069e5efcb7d0d912
Fix reader_ops_test.py in python3 by doing byte/string conversions. Change: 116037950
A. Unique TensorFlower
nobody@tensorflow.org
1,456,865,328,000
55be24524167c1d1879dba64b492c33c82244d1f
Fix bitrot in doc generator and run at head. Change: 116040450
Martin Wicke
wicke@google.com
1,456,866,334,000
b76812c8a84c0efd45342121bfc1cff2b6bb1051
Changes to reduce code size (especially useful for mobile): (1) Don't inline KernelDefBuilder::TypeConstraint (2) Don't inline AttrSlice constructors (3) Don't inline OpKernelConstruction::SetStatus (it's only called on error paths anyway) (4) Moved the slow path of OpKernelContext::record_tensor_reference into ...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,866,394,000
dd0a68cb77167890ac5a939020cdf72b234c395e
Minor doc fix for tf.constant: more precise about `shape`. Change: 116043312
Zongheng Yang
zongheng.y@gmail.com
1,456,867,730,000
f16aabc6d9c1f11d1b2c393544999d1ac960a80b
Adds GPU kernel for gather ops. Change: 116048950
A. Unique TensorFlower
nobody@tensorflow.org
1,456,869,874,000
bcf0d9ea050a1056c9c40a11dcb078c9a022e9dd
Make a better error message for restoring from an invalid file pattern. Change: 116049226
Martin Wicke
wicke@google.com
1,456,869,985,000
40ccc0aa632ebbd3b34f2f5da5f1ed576d50f57a
Improved the performance of outer reductions Change: 116063261
Benoit Steiner
benoit.steiner.goog@gmail.com
1,456,875,826,000
3f8a7616452346ef4de5a0da7b79fdb1cc8d6594
Rollback of "Adds GPU kernel for gather ops." Change: 116064181
Vijay Vasudevan
vrv@google.com
1,456,876,181,000
ec1403e7dc2b919531e527d36d28659f60621c9e
Add optional comprehensive logging of memory allocation/deallocation events. When enabled, the following events are recorded: The start of a step, with the numerical step_id and a textual handle describing the step. A Tensor allocation, including the step_id, the name of the OpKernel, the data type, shape, allocation...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,876,697,000
201b37589f0dbbdf938eaf5c0d1c40d416032698
replaced an unused parameter with `_`
ninotoshi
ninotoshi@users.noreply.github.com
1,456,747,267,000
773cb65fa89065126c5f640a8a61421509ad4ea7
Merge pull request #1331 from ninotoshi/anonymize_unused_var.
Vijay Vasudevan
vrv@google.com
1,456,878,954,000
8fa1b728b6649231cfcd9dee7138579812341b32
Add batch_norm_op.cc/h to Android extended operator filegroup. Change: 116067352
Andrew Harp
andrew.harp@gmail.com
1,456,877,742,000
c85167b07981bbe6cec3cacf2c692dcc5f5537fa
Remove .pyc file and add *.pyc to .gitignore
Yuya Kusakabe
yuya.kusakabe@gmail.com
1,456,499,521,000
ca7cb4e5ef7f6af36ea9e51561587e5ded42bbef
Merge pull request #1301 from higebu/remove-python-binary.
Vijay Vasudevan
vrv@google.com
1,456,880,896,000
b8ff42a4b3c0bbc4eadaa5ee5cd1b10adb58de7e
Merge commit for internal changes
Vijay Vasudevan
vrv@google.com
1,456,881,175,000
64017d18a9b7d2d2b1d39392bf925cd57bdf6fc9
Make the image, nn, and train packages only expose documented symbols. Change: 116070650
A. Unique TensorFlower
nobody@tensorflow.org
1,456,879,164,000
2ababd9c11a60ba4293aa0a3bffb4d68fc9aa89b
Adds squared-loss / linear regression to SDCA. Change: 116078284
A. Unique TensorFlower
nobody@tensorflow.org
1,456,883,054,000
115d41f35594a4df6cceb681d1bbf802f3a1ce10
TensorFlow: add "colocate_with" feature that uses _class attribute. Replace existing library functions that use device for colocation to use the internal library mechanism, which does a better job of ensuring colocation when devices are not provided. In other words, this allows colocation to be specified in a portabl...
Vijay Vasudevan
vrv@google.com
1,456,883,599,000
d25821376669ee193363d6098e127096029110af
TFRecordWriter explicitly calling close in mnist example.
e3
eerwitt@gmail.com
1,456,083,847,000
99952d68444cd2b08e88af972bd9eb7034fbc1e7
Merge pull request #1226 from eerwitt/mnist-convert-close-writer.
Vijay Vasudevan
vrv@google.com
1,456,885,179,000
4a8e97fb06817a45dd7a211838b2f49b8c12e6f3
Remove internal doc link and addressed TODO. Change: 116079599
Josh Levenberg
josh11b@tensorflow.org
1,456,883,778,000
2c40274367bfdcd226bc73d4515045ece2bf5de5
Fixing a subset of Python3-breaking lines currently in tensorflow Change: 116088162
A. Unique TensorFlower
nobody@tensorflow.org
1,456,888,778,000
9d8ea873db7246656c21f5860b53e34f14dd2841
Expose tf.contrib.util.stripped_op_list_for_graph C++ and Python use two different op registries, and in rare cases they can actually be different. Thus, we need both functions available. Also fix both Python and C++ to handle arbitrarily nested functions. Change: 116142089
Geoffrey Irving
geoffreyi@google.com
1,456,935,938,000
93587a2f3181d8c7464015e2ceb33c9ede532f54
Fix typo in relu_op.h. Change: 116147310
Zongheng Yang
zongheng.y@gmail.com
1,456,939,517,000
d3e8e04684fabf5e22d5993845f910e5d366bacb
Use java_generate_equals_and_hash=true for Features and Examples. Without it, generic reflective implementations of equals() and hashCode() are used, which are much slower. Change: 116150839
A. Unique TensorFlower
nobody@tensorflow.org
1,456,941,553,000
c3c40b7a8950bba876b92456824e5b6a9505c0a7
TensorFlow: fix OSX build Change: 116152701
Vijay Vasudevan
vrv@google.com
1,456,942,419,000
f07234db2f7b316b08f7df25417245274b63342a
Merge commit for internal changes
Vijay Vasudevan
vrv@google.com
1,456,942,751,000
5c378b0bb49f88538a115c4e3754fa3551da8c9a
Fix sync issue. Change: 116161343
Martin Wicke
wicke@google.com
1,456,946,534,000
96bd673708cd7f96c0b805bc256d651f8f016b62
Change TensorBoard Typescript version to 1.8.0 Change: 116164861
Dan Mané
danmane@gmail.com
1,456,948,269,000
431a2dfa8feece5362f732d7307f21f2ee70d0b6
TensorFlow: fix duplicate definition of BuildGraphOptions, fixes bug on mac build. Change: 116165342
Vijay Vasudevan
vrv@google.com
1,456,948,518,000
eda89e930cfcbd992ecacafd40267d733e2153dc
Merge commit for internal changes
Vijay Vasudevan
vrv@google.com
1,456,948,796,000
0b93524659052624f7c6437db2f0edc44bbb2644
Speed up graph serialization to pbtxt by a factor of 4.5-5x (tested on large graphs) The speed up comes in that __str__ uses the underlying c++ implementation to serialize, as opposed to the python one. Change: 116166881
Dan Smilkov
dsmilkov@gmail.com
1,456,949,321,000
3212eb3c874452ea928e23246e524b147f31cd15
Filter graph on server-side before serving it to the client. Updated the HTTP endpoint (and the doc) so the user can control filtering using the query parameters `limit_attr_size` and `large_attrs_key`. For now, the filtering includes removal of values of large attributes (size > 1KB). This significantly reduces the ...
Dan Smilkov
dsmilkov@gmail.com
1,456,950,322,000
80ddfbf11f63cf5d55ad16f4bc871cdd18ced3da
Add shape verification for `tf.nn.sigmoid_cross_entropy_with_logits()`. Fixes #1343. Change: 116175143
Derek Murray
mrry@google.com
1,456,954,429,000
39b59eb233e866da93c684bb9444b526f9286c29
Add license to log_memory.cc Change: 116178787
A. Unique TensorFlower
nobody@tensorflow.org
1,456,956,232,000
961ecf2a219494658f7f1c9acd32d6168efde43d
Mandating that the compiler cannot elide the copy to a local variable, in case there is unlocked concurrent access to the source tensor. No noticeable speed changes for scatter (it should not affect the asm output in any case I can think of; it's designed to prevent pathological corner cases.) Change: 116181195
David G. Andersen
dga@google.com
1,456,957,271,000
32ff58cd7b607e17c4182195729a07a5bf5bc7e0
We don't use tsd anymore since we migrated to typings Change: 116184216
Dan Smilkov
dsmilkov@gmail.com
1,456,958,657,000
a7298e47d9bc591527118ba7a882073d84e08599
Remove extra LICENSE files. Change: 116188107
Josh Levenberg
josh11b@tensorflow.org
1,456,960,554,000
60a315e352756fe6d42a20d806e1d9a8676fdf60
Make the contrib.layers, contrib.util, and python_io packages only expose documented symbols. These names are removed: tf.contrib.layers.initializers tf.contrib.layers.layers tf.contrib.layers.loss_ops tf.contrib.layers.python tf.contrib.layers.regularizers tf.contrib.layers.summaries tf.errors.OpError ...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,961,574,000
3455db797addd0843db4d00db65f49b139cfc5aa
Fix name in error message Change: 116193591
David G. Andersen
dga@google.com
1,456,963,474,000
b3a193de22c103a3e5189d5f548c1f9a9d10e3a0
Adds an in-process gRPC server with support for clean shutdown. This CL refactors the `grpc_server_lib` library such that it exports a `ServerInterface` class with `Start()`, `Stop()`, and `Join()` methods; and a factory method that takes a new `ServerDef` proto for configuring the server. Also fixes a bug in "master...
Derek Murray
mrry@google.com
1,456,964,096,000
f4b58903760a887489a407c514270f63823578da
Ensure that indices are validated in the face of asynchronous updates to the index tensor. Don't invoke Eigen functions on empty tensors. Change: 116198604
David G. Andersen
dga@google.com
1,456,966,142,000
868aa1e04a7574e74b93fc693c8cfcd62a5037eb
TensorFlow: reduce verbose logging. Fixes #1360, #1341, and more. We now only log the status at the server during executor failures if VLOG is 1. Similarly, do not LOG(INFO) when a dequeue/enqueue op is cancelled because the queue is closed. Change: 116198742
Jianmin Chen
goog.jmchen@gmail.com
1,456,966,233,000
1579f61b5cd1ad39c6a32d641a736e0c7e837027
Speed up SparseXentLossGenerator by eliminating a call to non-vectorized std::exp. This speeds up the ptb_word_lm benchmark by ~7% on CPU. % blaze build -c opt --copt=-mfma --copt=-mavx third_party/tensorflow/models/rnn/ptb:ptb_word_lm.par % ./blaze-bin/third_party/tensorflow/models/rnn/ptb/ptb_word_lm.par --data_path...
A. Unique TensorFlower
nobody@tensorflow.org
1,456,966,753,000
56808f9f19f1da45f53eac4f5c029840be6169a0
Merge summary_ops into logging_ops (both of which were small). Change: 116199874
Josh Levenberg
josh11b@tensorflow.org
1,456,966,818,000
643edf3db6082170bc4f7a5cf8eeb51875890b54
Rollback of "Adds GPU kernel for gather ops." Change: 116220950
A. Unique TensorFlower
nobody@tensorflow.org
1,456,987,637,000
5c8da7f82dbd0945f9599129cd6d55dc100eb689
Merge attention_ops (which only has one op) into image_ops. Change: 116256253
Josh Levenberg
josh11b@tensorflow.org
1,457,024,528,000
416390ed7adad906cf71857c620dd52ee3a79667
TensorFlow: allow importer to rename colocation binders when users import graphs using name scopes. Change: 116259458
Vijay Vasudevan
vrv@google.com
1,457,026,586,000
9218261e6b321445a988217c8e20ee3140a9dc16
Fixing a Python-3 bytes-string incompatbility in ops.py Change: 116259719
A. Unique TensorFlower
nobody@tensorflow.org
1,457,026,718,000
270b010b6d51ada90cbd9f7790b5dab71ce58bf7
Refactoring of SDCA related code for testability and addition of various sanity checks. Change: 116261455
A. Unique TensorFlower
nobody@tensorflow.org
1,457,027,792,000
ad392a422dffa4402f77c673feed9d184fbf0b83
Internal visibility update for contrib. Change: 116261711
A. Unique TensorFlower
nobody@tensorflow.org
1,457,027,942,000
7395855e6397eb15c3ce7fd91619eef6d292ae6c
Turn on exclusive mode for grpc session tests to prevent TOCTTOU port-binding races that cause test timeouts. Partially addresses #1363. Change: 116262977
Vijay Vasudevan
vrv@google.com
1,457,028,529,000
f16ee4b4c1cc2054bb6a640632806f11ff720e3f
Rollback of "Turn on exclusive mode for grpc session tests to prevent TOCTTOU port-binding races that cause test timeouts. Partially addresses #1363." Change: 116264279
Vijay Vasudevan
vrv@google.com
1,457,029,231,000
3871973b302b3c8357a4984fb37d934cf4633f21
Merge commit for internal changes
Vijay Vasudevan
vrv@google.com
1,457,029,476,000
e742b6d6ae3aba2c226a3f6e836910a03e218819
added libtensorflow_cc.so build target
Memo Akten
memo@memo.tv
1,456,922,497,000
589163da455b22aec078a024c131e8d72dfbb8a0
Merge pull request #894 from memo/master.
Vijay Vasudevan
vrv@google.com
1,457,038,644,000