hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
0279b3e6107d99df322afd09ca9211ccc9cbd3bf | Run gendocs
Change: 114262156 | Geoffrey Irving | geoffreyi@google.com | 1,455,058,097,000 |
fabbce38898572439852285388bbca2e19d5f412 | Fix all but one Python 3 test | Geoffrey Irving | geoffreyi@google.com | 1,455,049,466,000 |
9beac034d95164f5c8a4176ed9737760ed30907f | Merge pull request #1036 from girving/py3. | Martin Wicke | wicke@google.com | 1,455,070,385,000 |
de4441ef64132becae92cfc8c6505b71cfcfd654 | Improvements to print_build_info.sh and pip.sh
This is in response to @jendap's earlier comments. Using the $()
subscript syntax.
Also, making --upgrade default option for pip
This is addressing the issue that if --upgrade is not used, preexisting
installation of TensorFlow PIP on the system will make the install-test
results obsolete. | Shanqing Cai | cais@google.com | 1,455,031,636,000 |
a6fdd24cefafe8e2109f133bb89b1138ca538f11 | Merge pull request #1031 from caisq/master. | Martin Wicke | wicke@google.com | 1,455,070,458,000 |
86462105f53a1d0ef1eb7f6503ffb247f6e52db0 | + Comments for mnist example | terrytangyuan | terrytangyuan@gmail.com | 1,455,081,102,000 |
edefc77002a392f70eb79fca57daa35e4df934e7 | Improvements to TensorFlow android rules
Change: 114273085 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,063,853,000 |
7218dadf0de57abc426d6efa3902dc5dd9179c86 | Enables java_multiple_files for all tf protos, and sets the
outer class name consistently to <FileName>Protos. Also specifies
the java namespace as org.tensorflow.*.
This enables compiling tf protos with nano proto for Android
(which currently does not work because of file/message name
clashes) and follows the same convention for proto3 as used by the
API platform.
Change: 114279703 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,068,256,000 |
bc918afeb7e1768bd38563f980e3ec873f027468 | Add one-to-many multi-task sequence-to-sequence model.
This model has single encoder and multiple decoders which can be trained to produce different outputs for different tasks. Multi-task learning can help improve performance of the main task. Reference can be found at http://arxiv.org/pdf/1511.06114v2.pdf.
Change: 114283575 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,071,515,000 |
a4142f1289fea568b9fe8c0080bf4cdffa5a03ad | Includes ops used by user-defined functions in _stripped_op_list_for_graph.
Change: 114294037 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,084,868,000 |
65e48e88e17548e3116c3750dfd0b83c428e20ec | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,455,118,694,000 |
634993e3fd3b9a8a6b7f0cb9531e0fae6154cf52 | Sorting now runs alphabetically in tensorboard image grid. | panmari | smmuzi@gmail.com | 1,455,093,251,000 |
ca49d4de28aa110af697f153cdaa8158e0040177 | Merge pull request #839 from panmari/sort_runs_image_grid. | Martin Wicke | wicke@google.com | 1,455,120,768,000 |
880011919c7b113802eb23c54ed5b8e2218e4416 | Fixed tensorboard installation for pip
Broke when backend was moved to subfolder. See also #1000 | panmari | smmuzi@gmail.com | 1,454,948,560,000 |
7b6d2df2e0bdce552ebfdd2e88908e24de10db6d | Merge pull request #1020 from panmari/fix_tensorboard_backend_pip. | Martin Wicke | wicke@google.com | 1,455,121,677,000 |
56d8c5a85520d50ba447feb6028deecf1a420687 | Update os_setup.md with python3 documentation
Adds the python3 documentation for: "Test the TensorFlow installation" | Alexander Rosenberg Johansen | alexander@herhjemme.dk | 1,455,123,387,000 |
02e305c69c41e6a15d800e0514646916adc57895 | Merge pull request #1018 from alrojo/patch-1. | Martin Wicke | wicke@google.com | 1,455,123,412,000 |
d32b1095345cab195e6b64f01a132afa971b18f7 | fixed a typo | ninotoshi | ninotoshi@users.noreply.github.com | 1,455,102,927,000 |
033060f97fe760e024ab6b8fba3ce12202b22065 | Merge pull request #1039 from ninotoshi/master. | Martin Wicke | wicke@google.com | 1,455,123,530,000 |
211df3939c2e6216f2505ff6eb5e9423cc0cbb56 | Fixing a recently-submitted Python3-breaking line
iteritems() --> items() please | Shanqing Cai | cais@google.com | 1,455,138,517,000 |
5db4eb1bde4fd015f21b950df60ea0061c595563 | Merge pull request #74 from tensorflow/neural-translation
Continue working on neural translation model #9. Added language model example #64 | Illia Polosukhin | ilblackdragon@gmail.com | 1,455,139,717,000 |
e27aa15ed7fd4f3d735bf7a8a29b97fdede710bd | Add usage of batch norm in conv test and fix usage of is_training collection | Illia Polosukhin | ilblackdragon@gmail.com | 1,455,141,412,000 |
aeffe6bdd1722995382f1b3d37b23c26dd26e606 | Run gendocs
Change: 114328737 | Geoffrey Irving | geoffreyi@google.com | 1,455,121,411,000 |
2c13163b3fe88b1815f4010e261b542489487cf8 | Fix reduction op to honor host memory constraint.
Change: 114337745 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,127,154,000 |
60634dcbf8b226edf03c56485f253aad30f967c5 | Marks host memory for a few more ops.
Change: 114338500 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,127,511,000 |
d77ef35e133275c7d393d92e219276bb9eb47cad | Fixed the forward redeclaration of Eigen::ThreadPoolDevice and Eigen::GpuDevice
Change: 114341222 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,128,782,000 |
cab186e7a6128d5f8ea9f541afe05c173fed8a60 | Updated foldl, foldr, and map to work with TensorArray.
Change: 114350923 | Yuan Yu | yuanbyu@google.com | 1,455,133,546,000 |
18297126c0bc13c7045841f5a7a99f3da68176f4 | Fix tf.test for PEP-8 and document
tf.test now has appropriate snake case function names (get_temp_dir and
is_built_with_cuda) and has normal toplevel module documentation.
Also fix a bug in make_all.
Change: 114351269 | Geoffrey Irving | geoffreyi@google.com | 1,455,133,714,000 |
64e83b104a5dec6f47c77c7572f3e2a1a7dbc2f6 | Avoid a virtual function call to the real-time clock on the fast path
for every allocation.
Change: 114352030 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,134,097,000 |
c215dd646f701f5fd9e4d36f744ca9cd92faef1f | Splitting tensorflow android standard ops into two modules to enable separate compilation.
Change: 114356795 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,136,881,000 |
6edf204b65d4c9663484135a1634303b441be968 | Basic server tests for TensorBoard.
These tests start generate test data, start up a TensorBoard server, and then
make requests against it.
Change: 114358333 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,137,756,000 |
9934b9c6841ee56368c7cb1d10053f68a02680ba | Expose the ability to store and retrieve a datatype from the TensorShape
representation to the Tensor class, and use this ability to store the
DataType in the extra byte that is unused inside the TensorShape
representation. This makes sizeof(Tensor) == 32 bytes on a machine
with 8-byte pointers, instead of 40 bytes.
Change the value field in AllocatorAttributes from uint32 to uint8, and then
rearranged the order of the fields in ExecutorState::Entry so that this fits
in the same word as the has_value boolean. After the Tensor change above,
Entry was 72 bytes, and this change makes it 64 bytes.
Change: 114358788 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,137,982,000 |
b2757fb5797252bb1ef12bbf04be796cc83000cc | Switch label_image to use TopKV2 instead of TopK
Fixes #1040. I need to file a further bug so that deprecated ops are removed
from the C++ API entirely, similar to what we do in Python.
Change: 114360963 | Geoffrey Irving | geoffreyi@google.com | 1,455,139,295,000 |
ea92856334a74519055a4e22aba7ceee5dd039a2 | Add versions to checkpoints
Checkpoints now have a version scheme analogous to that for GraphDefs. We have
no plans to ever deprecate a checkpoint version, but it's good to have the
scheme in place in case we need to.
Change: 114364388 | Geoffrey Irving | geoffreyi@google.com | 1,455,141,167,000 |
bc6f617bd636c4e371d7c03d153e9ce9b8b0e3c3 | * Change gpu_count to gpu_device_enabled.
* Switch to have the BaseGPUDevice constructor to enable GPU support, instead
of relying on the entry point.
This make it possible for TensorFlow to use pinned memory for GPU/CPU memory
copies for all entry points.
Change: 114364920 | Xiaoqiang Zheng | zhengxq@google.com | 1,455,141,435,000 |
7760ce56fc3ab4ab8cdc408e29d8ad8b539c417e | Get rid of some import cruft.
Change: 114374558 | Josh Levenberg | josh11b@tensorflow.org | 1,455,145,992,000 |
565f3e835294d25d6939fc8eb01dff5489798773 | Comment typo: vetors -> vectors
Change: 114374769 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,146,099,000 |
fc547c1289793d24ad0da06b4ce20a8da7f34e64 | Update ops.pbtxt to include typo fix.
Change: 114377444 | Josh Levenberg | josh11b@tensorflow.org | 1,455,147,581,000 |
18cbcdfaabf58588675cf9c29e2b9ec84602422e | Add an optional argument to model_with_buckets to get per-example loss and
correct the use of conditionals in seq2seq functions to make them efficient.
Change: 114377454 | Lukasz Kaiser | lukaszkaiser@gmail.com | 1,455,147,586,000 |
4c8a0f19ee5eae260d54250d1ed6f328a5a7831a | Fix bit rot in random op benchmarks
1. RandomParameters is now TruncatedNormal.
2. Shapes can't be scalars anymore.
Change: 114377586 | Geoffrey Irving | geoffreyi@google.com | 1,455,147,660,000 |
3752fcf86279a15aeb85c83c37a211a6cc99f042 | Adding argmax and equal_to to android extended ops.
Change: 114378906 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,148,460,000 |
68b9c8a3e3cf26b6dcf17ef509dd14388793ce65 | Changed from using general tmp folder to per-test tmp folder. | Illia Polosukhin | ilblackdragon@gmail.com | 1,455,150,384,000 |
257f98142e9b13c4392be3f6820928b8b573eee1 | Merge pull request #1044 from caisq/fix-python3-4. | Vijay Vasudevan | vrv@google.com | 1,455,150,804,000 |
357eeee4123d32d5943afd706bcdf6eaae317583 | Update README.md | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,455,161,526,000 |
648d38d0354a07bddd4786423c10006ad1bc862f | Restore lost changelog entry | Martin Wicke | martin.wicke@gmail.com | 1,455,165,113,000 |
f27a4d90843fc56e492f7534f6f7c6b722e17cd6 | Fix typo in RELEASE.md | Martin Wicke | martin.wicke@gmail.com | 1,455,166,684,000 |
5181df88852a779b84dfe3a6dbe5c0f3515145f4 | Completed the cleanup of the TensorFlow local copy of the Eigen codebase | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,166,548,000 |
f3fe65a3cda29cb9df7386ee7f0305a1dcaa93d9 | Merge pull request #1053 from benoitsteiner/master. | Martin Wicke | wicke@google.com | 1,455,170,427,000 |
d1a0ea5ae05d7c1a88d7ec4d3d139962d5b16338 | Fix new typos in tensorflow/core. | Dongjoon Hyun | dongjoon@apache.org | 1,455,147,986,000 |
f389c08e77662e395eb4af0d915f18467dcc8462 | Merge pull request #1048 from dongjoon-hyun/fix_new_typos_in_core. | Martin Wicke | wicke@google.com | 1,455,170,891,000 |
0568469ac8582058919f05a50f70c3ff83912b0c | Fixed "filer" type in variable_scope how to
Changed "filer" to "filter" | rizzomichaelg | rizzomichaelg@users.noreply.github.com | 1,455,124,247,000 |
9f87760a837ec842de00a50f5688e2684110899d | Merge pull request #1042 from rizzomichaelg/master. | Martin Wicke | wicke@google.com | 1,455,171,120,000 |
3416531d4aa0f6e394d3e174b9da9066983dab24 | Update resources page | Martin Wicke | martin.wicke@gmail.com | 1,455,174,217,000 |
c1f08455d8e650f6dd96a0330111b9b573d059dc | Merge pull request #1054 from tensorflow/update-resources. | Martin Wicke | wicke@google.com | 1,455,174,239,000 |
8badb8c8ea08dff8fba6c4177cc4f78972fdaed1 | removed signed unsigned comparison warnings for Move in inlined_vector.h | chemelnucfin | alan.c.wu@gmail.com | 1,455,174,430,000 |
c4207090d68151fb967672a90ea6d574e62cfba0 | Merge pull request #1050 from chemelnucfin/fixcomparison. | Martin Wicke | wicke@google.com | 1,455,174,459,000 |
d584fe30abb333ade4ace1876d2c9711db21feb5 | Upgraded to the latest version of Eigen that provides better support for
embedded platforms and optimized implementations of the tanh function
amongst other enhancements | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,171,149,000 |
efb860cf11ff814b695ab64b20b206944f0287d8 | Merge pull request #1055 from benoitsteiner/master. | Vijay Vasudevan | vrv@google.com | 1,455,216,043,000 |
679bde447f94e31f5b5199728d1085496c1e15ad | Merge branch 'master' into patch-1 | jmtatsch | tatsch@gmail.com | 1,455,217,017,000 |
9043af8ab7e075f843adc37d95912dd7f9c62c66 | Save the graph_def corresponding to the current default graph instead
of a cached version.
Change: 114388651 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,154,462,000 |
fe398ae67c737198e8db9923312a4a764ccd6947 | Deleted unused constant
Change: 114397856 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,164,425,000 |
73b9dd18ce3017829edef4a5b4190a4f0579369c | Fixed several saver_test errors in open source and with python3.
Fixes #1035
Change: 114439527 | Sherry Moore | sherrym@google.com | 1,455,207,294,000 |
7415dfae93d4ab3e1f15f874bf7a42f82cf8b377 | Moves MemoryType inference code out of OpKernel so that it can reused.
Change: 114448861 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,213,975,000 |
402f5e94a9cfcc04c0f19171d6db6a5d1a48cc1f | Update generated Op docs.
Change: 114452187 | Josh Levenberg | josh11b@tensorflow.org | 1,455,215,483,000 |
dd353d81c05864b0c9175b8b81362e568cbf5e66 | Use six.moves.http_client instead of httplib.
Change: 114455261 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,216,970,000 |
88b0cb44a468ca8c26b20f008d3414a1474a3f8e | Force the executor to wait until queued operations have finished executing
before allowing the SINK Op to complete. This probably better matches the expected behavior, and avoids misleading output from benchmarks, as well as marginally improving the reported timing of some of our existing benchmarks, and reducing the amount of "burn-in" time needed before getting useful readings from benchmarks.
A common pattern in benchmark python code is:
start_time = time.time()
_ = session.run(tf.group(*target))
duration = time.time() - start_time
Since no tensor is fetched by session.run the call returns as soon as the SINK Op has finished its Compute method. With the current codebase this does not wait for GPU-queued operations to complete, and as a consequence steps that should execute sequentially are overlapped. By modifying benchmark_overfeat (in this case with batch size 2750) to print out timings for every step including "burn-in" we see the behavior, where between 5 and 6 steps are overlapped before backpressure kicks in (the direct session thread pool has 5 entries) and we see the steady state time stabilize:
2016-02-11 10:38:13.556584: step 1, duration = 0.005
2016-02-11 10:38:13.561401: step 2, duration = 0.005
2016-02-11 10:38:13.565904: step 3, duration = 0.004
2016-02-11 10:38:13.570100: step 4, duration = 0.004
2016-02-11 10:38:15.434168: step 5, duration = 1.864
2016-02-11 10:38:23.401597: step 6, duration = 7.967
2016-02-11 10:38:31.372718: step 7, duration = 7.971
2016-02-11 10:38:39.344853: step 8, duration = 7.972
2016-02-11 10:38:47.333486: step 9, duration = 7.989
2016-02-11 10:38:55.297039: step 10, duration = 7.963
2016-02-11 10:39:03.267716: step 11, duration = 7.971
2016-02-11 10:39:11.243412: step 12, duration = 7.976
2016-02-11 10:39:19.198777: step 13, duration = 7.955
2016-02-11 10:39:27.175440: step 14, duration = 7.977
With the change the output (same batch size) shows:
2016-02-11 10:35:24.421367: step 1, duration = 7.832
2016-02-11 10:35:32.269099: step 2, duration = 7.848
2016-02-11 10:35:40.118505: step 3, duration = 7.849
2016-02-11 10:35:47.946552: step 4, duration = 7.828
2016-02-11 10:35:55.802172: step 5, duration = 7.856
2016-02-11 10:36:03.637670: step 6, duration = 7.835
2016-02-11 10:36:11.505289: step 7, duration = 7.868
2016-02-11 10:36:19.331914: step 8, duration = 7.827
2016-02-11 10:36:27.162643: step 9, duration = 7.831
2016-02-11 10:36:34.994089: step 10, duration = 7.831
2016-02-11 10:36:42.850624: step 11, duration = 7.856
2016-02-11 10:36:50.707724: step 12, duration = 7.857
2016-02-11 10:36:58.574334: step 13, duration = 7.867
No steps are overlapped, so the first step reports meaningful timing, and as a bonus the overall reported step time is slightly less, presumably because we are able to use all 5 CPU threads in the thread pool for a step.
The maximum batch size before OOM is the same before and after the change.
Change: 114458853 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,218,818,000 |
8048088bfb82846078f8023bc6199e6424926624 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,455,221,584,000 |
dc78693f1300a83a56a03941c08e3e8dd8009b84 | Clean up references
Clean up references and add explicit links to pdf files for references
that are displayed on www.tensorflow.org | Jonas Meinertz Hansen | JonasMeinertz@users.noreply.github.com | 1,455,224,305,000 |
ed6f783690a35e54441881930d4291894bc03285 | Merge pull request #1057 from JonasMeinertz/fix-arxiv-links. | Vijay Vasudevan | vrv@google.com | 1,455,224,330,000 |
27eb320f9d78e51828fb80cdab90cbb76c6a43e5 | Fix a bug for an unavailable argument | Kenta Yonekura | yoneken@ieee.org | 1,455,205,813,000 |
7718d2a052042d6ee1d821c979661371e217f0d7 | Merge pull request #1058 from yoneken/bug-onehot. | Vincent Vanhoucke | vanhoucke@google.com | 1,455,230,692,000 |
c4801e3624dec02091009b40dd9c7e28aed526b2 | Clean up saturate_cast, test, and move to tf.saturate_cast
Change: 114470777 | Geoffrey Irving | geoffreyi@google.com | 1,455,225,780,000 |
62622d78c5cd45fbbe04497de856ff5424051678 | Fix doc string weirdness introduced by script
The doc string is still pretty confusing, but this is what it was before I
messed it up with my script.
Change: 114470999 | Geoffrey Irving | geoffreyi@google.com | 1,455,225,911,000 |
3c933761e40101952fdcc0896bb3de1c5654192b | Fix TODO addressing imports in gen_cc_md.
Change: 114471496 | Josh Levenberg | josh11b@tensorflow.org | 1,455,226,211,000 |
80e682562a202d5ed57c871aa76f0faec4061590 | Switch the slow path in matrix_solve_ls to using Eigen::CompleteOrthogonalDecomposition (COD), which I recently contributed to Eigen in https://bitbucket.org/eigen/eigen/pull-requests/163/implement-complete-orthogonal/diff
The advantage of COD over column pivoted QR is that it is able to compute the minimum-norm solution when the matrix is rank-deficient, which is usually the desired behavior and makes it consistent with the fast path.
Change: 114483303 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,232,451,000 |
0e28e705d377c554d2d750f69fa11126f51345a3 | Hardening the error checking related to memory types.
Change: 114484861 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,233,340,000 |
c1e0d8d469ed467f7421a5a528895feb096a4a07 | Avoid calling the default constructor of GpuDevice since it puts in motion a
lot of expensive stream executor machinery. Moreover the stream executor
initialization was done for nothing as the device always ends up being
reinitialized before its first use.
Change: 114488347 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,235,420,000 |
ad207ba7307e85d0d54b1cafe3ffb76db7f2c4d4 | Stop using deprecated imports in rnn.models (rnn, seq2seq, etc).
These are all now in the tf.nn namespace; just call them directly.
Change: 114488673 | Eugene Brevdo | ebrevdo@gmail.com | 1,455,235,580,000 |
4f8670d47c9dbfb210a5797d83a2d8b01f94845d | Rename "demo_from_server" to "serialize_tensorboard" and improve it.
serialize_tensorboard.py is a script that saves all of the data out of a TensorBoard instance to disk, so that the TensorBoard frontend can be developed against this saved data without needing to run an actual TensorBoard instance. Also, this data is used to power the demo tensorboard on tensorflow.org.
This CL changes serialize_tensorboard in the following ways:
- it now takes a logdir directly (like TensorBoard), rather than requiring that a TensorBoard instance be turned on first.
- it now uses flags and platform.app.
- it now saves data to disk in exactly the same path structure as TensorBoard's URLs, meaning that there is no need for a special demo router in the TensorBoard frontend.
Change: 114493436 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,238,814,000 |
62b0b4ac302bb49dfebb054a26c044bf41a65112 | Add a tensorboard script that generates fake event files so we can use them for consistent debugging.
Change: 114493731 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,239,010,000 |
7ac6fdf592c246c20720023a153f0502a61091b7 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,455,240,184,000 |
174fd60ebe78b22288c6e9b6c079c3ebb58f56d0 | Optimize TensorSliceSet overlap checking.
Maintain 'hull slice' which covers all existing slices. If a new slice doesn't
intersect with the 'hull slice', then there is no need in checking all existing
slices.
This optimization reduces O(N*N) time to O(N) when slices are being added
consequently.
Change: 114498099 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,242,742,000 |
4392eee66fe9e452e5b9ce8188347a58476f56f3 | Add missing std namespace to min and max calls in tensor_slice.cc.
Change: 114500524 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,245,292,000 |
01806e89e5dd141ae456bfee2ca0e799fb1aa32d | Add the complementary operations TensorArraySplit and TensorArrayConcat.
These operations are each other's gradients.
Change: 114503529 | Eugene Brevdo | ebrevdo@gmail.com | 1,455,248,886,000 |
6b884f0bc592b146f64b408a684937b62bc6b6d3 | Added tool to inspect checkpoints.
Change: 114506562 | Sherry Moore | sherrym@google.com | 1,455,252,933,000 |
69e4339de96e1a0af05aee300cdcdd3ea6ec653d | Have tensorboard/scripts/generate_testdata.py import google3
Change: 114508302 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,255,258,000 |
3f8748554355a06644ece0f9832d5986a657f11e | Added SUMMARY_OP to GraphKeys.
END_PUBLIC
Made necessary changes to supervisor to handle restoring from saved meta graph.
BEGIN_PUBLIC
Added SUMMARY_OP to collections.
Change: 114510124 | Sherry Moore | sherrym@google.com | 1,455,257,606,000 |
621354221a17885486ced9ce463b15fbe38536e9 | Link with -lm.
Framework calls ceil somewhere deep inside.
Change: 114539499 | Manjunath Kudlur | keveman@gmail.com | 1,455,296,702,000 |
66301e45cc131c17fe89f17359519bd0325e7664 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,455,297,079,000 |
ca7efd72d111cd6876c4c0b1279b3ba0b61f67bb | Allow encoding of gray-alpha PNGs.
Change: 114540382 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,297,351,000 |
fd546856cc2ee68c6487d4a534a78bf0b8a933cc | Turn off the background polling loop by default in all
gpu_event_mgr_test cases. Race conditions are possible between the
background polling loop and the actions of the test which make
consistent measurements very difficult. The tests now always rely on
the test helper PollEvents which is synchronous.
Change: 114540935 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,297,788,000 |
125b8058c0b815c7b1b2663ff2d7568d81c08313 | Add a bitcast op for changing the type of a tensor without copying its buffer
Fixes #655.
Change: 114542161 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,298,695,000 |
985528bd85c22b7b6dbced958e0b8a12c5bdffcb | Use gfile wrapper instead of direct file access in mnist input builder.
Change: 114545370 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,300,634,000 |
ee544380869053c6826bd04c4d3ff4ce1ab2a2da | Adding a NodeRadar class for monitoring which charts are in view.
Change: 114547575 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,301,904,000 |
5e64274954c32f0959dfe14fc39cf981e3d3134c | To statically check if a fetch can be computed with the current feeds, we keep a copy of the graph in the case of partial run. We need to keep the graph in ExecutorsAndKeys rather than RunState.
Change: 114556076 | Yuan Yu | yuanbyu@google.com | 1,455,306,740,000 |
20d7853396c6fde4c2f273f252c022fad2674062 | Simple end-to-end benchmark for batch normalization.
On my crappy desktop:
Forward convolution.
shape:4/3 #layers:5 python:False scale:True train:False - 1.393189 secs
shape:4/3 #layers:5 python:True scale:True train:False - 1.644035 secs
Forward/backward convolution.
shape:4/3 #layers:5 python:False scale:True train:True - 3.938029 secs
shape:4/3 #layers:5 python:True scale:True train:True - 4.542710 secs
Forward fully-connected.
shape:2/1 #layers:10 python:True scale:True train:False - 0.096353 secs
Forward/backward fully-connected.
shape:2/1 #layers:10 python:True scale:True train:True - 0.218981 secs
Change: 114559841 | Vincent Vanhoucke | vanhoucke@google.com | 1,455,309,235,000 |
ef2c2a189e5015c04619eb6bc30e4d073eadfa2a | Avoid unecessary duplication of the Eigen codebase.
Change: 114560374 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,455,309,623,000 |
8fce27bc797582c573c232b9b9328cef4c642048 | Fix int32-overflow error in shape computation.
I got the following error message before tracking down this bug:
InvalidArgumentError: Input to reshape is a tensor with 2764800000 values, but the requested shape requires a multiple of 2764800
Change: 114562418 | A. Unique TensorFlower | nobody@tensorflow.org | 1,455,311,031,000 |
fa112c3d97a3bacb199fc0f288d24986ee5405ec | TensorFlow: fix android OSS build to remove stray rule
Change: 114565136 | Vijay Vasudevan | vrv@google.com | 1,455,312,804,000 |
68e8b0f1e02f1e0e10f4fdd689ede80f973e2756 | Fixes issue with loading external tensorboard assets | Daniel Mané | danmane@google.com | 1,455,312,934,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.