hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
16a671c0c5977f3d04c36811a72412cd0d74296f | Execute TODO to make #include more precise.
Change: 111457566 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,498,000 |
96689166e9dbc2c740e5013a49673cd38e8683d1 | Remove #include of mem.h from port.h, instead including it
just where it is used.
Change: 111459203 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,506,000 |
be39348393d977ff37d3eab2220253bc0b1d0c8d | Explicitly setting shape in grayscale_to_rgb and rgb_to_grayscale.
Fixes shape inference following a previous change that generalized these two members of image_ops. The more general implementation, however, does not allow shape inference. This change explicitly fixes the shape with set_shape, following array_grad._TileGrad.
Adding tests for expected shape inference behavior.
Change: 111459347 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,517,000 |
231c853a3ad5ca8ae7564fcb9c6a3d28175d2309 | Change: 111459848 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,525,000 |
d0dfcdbd7bf605f5878f1b3aa2aacd5fcf001e32 | Change: 111461651 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,542,000 |
0ce489f4e0c3b12d4dce1b2d642f48f45797774a | New GraphDef version 4: TensorFlow is now scalar strict
The kAllowLegacyScalars flag is now gone. Instead, scalar strictness now
depends on the GraphDef version: we are lenient below 4 and strict with 4
and above. All new graphs should use version >= 4.
The internal Google version of tensorflow is now as scalar strict as the
open source release. Note that outside of PLATFORM_GOOGLE, this change has no
effect, since the code was already scalar strict.
Change: 111467133 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,550,000 |
21d78082a727a4a3d95959611ee6a0877a260591 | Print mnist convolutional per-step times
Change: 111468076 | Yaroslav Bulatov | yaroslavvb@google.com | 1,452,113,557,000 |
4519ff87fc21538079be670a9a4098ee890e1d0b | Update docs.
Change: 111469707 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,566,000 |
98b50f742d6e15dd93b12978eac199f3663e2c68 | TensorFlow: switch port_test.cc to include mem.h. Fixes OSS build.
Change: 111475504 | Vijay Vasudevan | vrv@google.com | 1,452,113,582,000 |
2c64510a149bbcd9a46fa9ba5ad13faa4d5e7d73 | Drop host_info.h from port.h and #include it directly
where needed.
Change: 111476610 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,590,000 |
d64612884d88a200f86ce4e602e49b8cbe0d4e39 | Move TF_ANNOTATE_MEMORY_IS_INITIALIZED from mutex.h to mem.h.
Update the one user of TF_ANNOTATE_MEMORY_IS_INITIALIZED to #include mem.h.
Change: 111478711 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,608,000 |
71677020b92898da52d26885ce269b104ec863b7 | #include third_party/tensorflow/core/platform/macros.h
directly so we can drop it from port.h.
Change: 111506630 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,616,000 |
252bb90ca2dd412ca2fd2908faf1a25d6ef618cf | #include third_party/tensorflow/core/platform/macros.h
directly so we can drop it from port.h.
Change: 111506675 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,625,000 |
e26f0f34018b7adb519e989b59a0462b08a93ea8 | In-embedding nodes (constants) can have inputs such as control dependencies. When embedding a constant into a node A, move the inputs of that constant into inputs of A.
Similarly, when embedding an output node into node A (e.g. a summary node), move the outputs of that summary into outputs of A.
Change: 111508145 | Dan Smilkov | dsmilkov@gmail.com | 1,452,113,634,000 |
b947bcc154db62e2306e8bf3edb1a94868edbbca | TensorFlow: Remove use of command_line_flags library in our usage. Soon
to be deleted completely.
Change: 111521876 | Vijay Vasudevan | vrv@google.com | 1,452,113,643,000 |
a1d4149a6eceebf2a5634b67e0c3a92566e7e566 | Change: 111522571 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,661,000 |
69e9ff4b8f97b870e613d97bffa117598007be1a | Fix parse_single_sequence_example to not fail when debug_name is None
Change: 111526769 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,670,000 |
911cec6b808e841fee7c742c85bac8c4b4a3113e | Change: 111526818 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,678,000 |
a327fa069cfe9ac54986f00586309124ed037926 | Allow slicing up to rank 6 tensors.
Change: 111527673 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,685,000 |
953660c5630b5d3bdf4567b68ba6e51929395ee5 | Fix build breakage from a typo.
Change: 111528530 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,693,000 |
3ae059ec4952493366510e4180773a2e2f398510 | #include third_party/tensorflow/core/platform/macros.h
directly so we can drop it from port.h.
Change: 111528649 | Josh Levenberg | josh11b@tensorflow.org | 1,452,113,702,000 |
9edd13a9c96edd9b7fdca71f694bb045f265ad66 | Rename deconv2d to conv2d_transpose and expose publicly
deconv2d is a misleading name for an operation which is exactly the transpose
of conv2d. conv2d_transpose is much better.
Fixes #256.
Change: 111529902 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,710,000 |
78e8dc03486d579e0ba6692849b9db72e4fbc1f7 | TensorFlow: include types.h to unbreak OSS BUILD
Change: 111531092 | Vijay Vasudevan | vrv@google.com | 1,452,113,718,000 |
fc909eda0059d6586e7c4bee7412116d8ef033fd | Update generated Op docs.
Change: 111535183 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,113,741,000 |
b5b11c4f84a6c80a3c373aa8c5fb093c4d58ea5f | Split out platform-independent stats collection code from tracing.h.
Change: 111535475 | Derek Murray | mrry@google.com | 1,452,113,749,000 |
5b2e90ae734c1c42c7b7c42db56df53b4c48c6a8 | TensorFlow: One more bug fix in example_trainer.cc for OSS build.
Change: 111535648 | Vijay Vasudevan | vrv@google.com | 1,452,113,759,000 |
b6acce62fb3e426dc4521fbf366762bb4c02d28f | Improved the performance of the contrast adjustment op by manually computing
the mean.
Change: 111535650 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,113,768,000 |
45f82c7a7675cb9fdef249b8a6e916d75d828385 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,452,114,556,000 |
d12b67833196d5508657d488cb8d96127419f2eb | Change: 111539073 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,118,104,000 |
2837487234aa11ac21a2c3b0aff33e673b72c58b | Change: 111542313 | Vijay Vasudevan | vrv@google.com | 1,452,118,124,000 |
d5fcafba94876c14057f8f928bb9276e285c9f94 | TensorFlow: move grad declarations from cc/ops to core/ops
since they aren't part of the generated API. remove cc_library
in py_library which breaks OSS build.
Change: 111542333 | Vijay Vasudevan | vrv@google.com | 1,452,118,132,000 |
fd26a93ee1799c78975a02c7bba58b3e52357fb3 | Don't expose tf.ops in __all__.
It was already undocumented, and now it's invisible. This required exposing
reset_default_graph and register_tensor_conversion_function as public.
Fixes #266.
Change: 111542779 | Geoffrey Irving | geoffreyi@google.com | 1,452,118,158,000 |
b09a56f462131f52bb49ca39b64334ed43051019 | Merge pull request #693 from frankyjuang/master
Fix file exists error | Martin Wicke | martin.wicke@gmail.com | 1,452,118,508,000 |
8f8530bf56e0abb6ce63e8c0fb99de76a0bc759b | Fix documentation typos in ops.
Since this PR does not include the fixes on generated docs,
the corresponding md files need to be regenerated after merging this PR. | Dongjoon Hyun | dongjoon@apache.org | 1,452,125,371,000 |
bed23441ff92cac70856c9ab555e7ef8043712b2 | New docs.
Change: 111545544 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,128,593,000 |
b517f837c55937528591c8c1ef895ac59774d8e1 | Avoid creating a temporary tensor when computing the contrast adjustment. This
helps improve the performance of the op when running on GPU.
Change: 111546411 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,128,601,000 |
0472586fa7d8ee6695765fe9a6b94fce5984fdfd | Fixed the adjust_contrast benchmarks: the op only supports floating point
values as input, so feeding it 8bit integers wasn't correct.
Change: 111548739 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,128,628,000 |
59cbb37188bb80e4a2738c616c50ba45cdfd96c7 | Adding more features to Android build:
- make tensorflow.Example and tensorflow.Feature protos available.
- make snapshoting and a few more operators available in android_extended_ops.
- if __ANDROID_TYPES_FULL__ is set, enable quantized types.
Change: 111551874 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,128,642,000 |
7b00fe3f2605491cf62d4e1b2818f375e79c8b8c | Improved the performance of the contrast adjustment op by a factor of 3.
Change: 111555964 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,128,652,000 |
816adeef1c6e9c48b31d3426b565a7851f7f8e39 | Refactor how TensorBoard url overwriting works.
Currently TensorBoard urls are changed by modifying a global variable (TF.Urls).
This creates racyness where we need to ensure that that global variable changes before the TensorBoard instantiates.
It also makes it impossible e.g. to have multiple TensorBoards on the same page using different url schemes.
This CL changes TensorBoard so it takes in a "router" object which specifies what the routes are.
It also adds a tf-tensorboard-demo component which is just like a TensorBoard, except it has a "dataDir" property that determines where to look for demo data.
Change: 111558450 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,128,673,000 |
25e634de6e9dae94d52440462d504e20786dc479 | Merge pull request #695 from jimfleming/shared_library
Add BUILD rule to create a shared library | Martin Wicke | martin.wicke@gmail.com | 1,452,129,675,000 |
82cc364bc969ae5dae21a2f5e6c6278e6b8d01c9 | Merge pull request #698 from benoitsteiner/master
Deleted the TF copy of the tensor code | Martin Wicke | martin.wicke@gmail.com | 1,452,129,746,000 |
45199e25a96c28209767dcf5158a642f9073a855 | Merge pull request #710 from dongjoon-hyun/fix_typos_in_ops
Fix documentation typos in ops. | Martin Wicke | martin.wicke@gmail.com | 1,452,129,821,000 |
5f791dee245a8d684624c4daa9cd326ab4ed8971 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,452,130,467,000 |
8d6a6c3d2209451e39d67e8445662627ca337358 | Use inspect to avoid virtual env custom site package pypa/virtualenv#228 | e3 | eerwitt@gmail.com | 1,452,030,346,000 |
3d48da3a8412061dd239e22ecec21ac5ede4f4cd | Using pack on decode_csv results to make example CSV file parsing run. | e3 | eerwitt@gmail.com | 1,452,133,502,000 |
9d5f623b9bbf3072443ecee6efac7bcd76ecb32d | Merge pull request #712 from eerwitt/csv-example-using-pack
CSV file loading example using pack | Vijay Vasudevan | vrv@google.com | 1,452,138,400,000 |
63bddb3efa7814f1451b4a190e56ff7ff856e380 | Merge pull request #703 from eerwitt/check-sitepackages-in-virtualenv
Test the TensorFlow installation location using inspect | Vijay Vasudevan | vrv@google.com | 1,452,138,638,000 |
b6ff1417eff4b3a95422d85b8e20a9bff026c090 | Merge pull request #70 from terrytangyuan/saver
Additional Options for Saver | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,147,299,000 |
089d99ba704b33fa95a0a17cad5227aa45d59621 | Moving extraction of pandas dataframe into io/data_feeder | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,148,708,000 |
135018dbe63d7d72201862d0c5e5878d1fa4a923 | Added more coverage in tests | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,149,686,000 |
398f209431b95671b6e48c4456297d175864c3f1 | Adding more tests to categorical vocabulary and not fitted errors in estimator | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,150,348,000 |
e4ae3aa1fe948ab5ccc6f69035e96d068d79a494 | More extensive comment about summary writer initalization | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,150,507,000 |
25bfa796420d88e9c8f5a902e41f2d02b0a027cc | remove training whitespace | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,151,715,000 |
4f31fb0dc02c3e9c2ac58e79abec57a7a544eef4 | Change: 111565458 | Manjunath Kudlur | keveman@gmail.com | 1,452,181,523,000 |
55d05620c3113a17e2b71bf612060e50f72325af | Autogenerated Change: Release TensorBoard at TAG: 5
Change: 111565559 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,181,531,000 |
1fbc7412ba95412eeae4189c7aee56877de36863 | Autogenerated Change: Release TensorBoard at TAG: 6
Change: 111566354 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,181,539,000 |
573fbda10163d4d12871f824acbfd1fb5778332f | Update assignment #5 with problem.
Change: 111574654 | Vincent Vanhoucke | vanhoucke@google.com | 1,452,181,547,000 |
4be49cce22726281a90851d12c99bfdbbeb2739c | Check the result of serializing a Tensor and return an error if it could not be serialized (most likely because it could not fit into a protobuf). This prevents segfaults when trying to load invalid protobufs.
Change: 111576943 | Andrew Dai | adai@google.com | 1,452,181,562,000 |
5d824f10cb811cfaf7988cb647ac553cb951eb86 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,182,199,000 |
492812c7b4885c331912995675bd8dd5bb2a95aa | remove another extra whitespace | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,195,137,000 |
ee403f003e4614baab6219bedb19f5c15a6a1225 | Ref #72: Update histograms for X and y only if they are floats | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,195,337,000 |
564abcc02f98ae83f8ae9969a7546b510efbbb94 | #include third_party/tensorflow/core/platform/macros.h
directly so we can drop it from port.h.
Change: 111612474 | Josh Levenberg | josh11b@tensorflow.org | 1,452,201,852,000 |
ded61708dbc7f5615119a3623c8fe7f112882e0f | #include "tensorflow/core/platform/mutex.h"
directly so we can drop it from port.h.
Change: 111613572 | Josh Levenberg | josh11b@tensorflow.org | 1,452,201,862,000 |
b1d45c972e0f1b600ea269f85fce63b79ef9135b | #include "tensorflow/core/platform/mutex.h"
directly so we can drop it from port.h.
Change: 111613643 | Josh Levenberg | josh11b@tensorflow.org | 1,452,201,871,000 |
af9b56881008f8b804cf93c0d99cb96357b19748 | #include "tensorflow/core/platform/mutex.h"
directly so we can drop it from port.h.
Change: 111613760 | Josh Levenberg | josh11b@tensorflow.org | 1,452,201,880,000 |
0389bf87d657784716924964b7b936d170a27b38 | Fix the image dashboard
Change: 111615936 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,201,900,000 |
68aab4180381054a6a87fb4b485569912674c54c | Autogenerated Change: Release TensorBoard at TAG: 7
Change: 111617377 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,201,916,000 |
f13442a8c6abadf61671ab70400c6ad287f4dcee | The default memory manager would delay 10s for any memory failure. This is not good for Cudnn scratch space allocation, which is only optional with performance gains.
Adding per-allocation attributes so BFC GPU allocator can avoid waiting for
memory failures for Cudnn scratch space allocation.
Change: 111617463 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,201,928,000 |
f3d46653e9780fe640d242b3e98712ba8569168b | Make sure new default values are set when loading old graphs.
Change: 111621107 | Pete Warden | pete@petewarden.com | 1,452,201,945,000 |
3ffa307e49e5b150934a71386194d7ed621e3e98 | #include third_party/tensorflow/core/platform/macros.h
directly so we can drop it from port.h.
Change: 111621646 | Josh Levenberg | josh11b@tensorflow.org | 1,452,201,955,000 |
6cc392e3b0989744c7b16248b19b48bcbe54fa6a | some linting fixes to changes brought in from the public.
Change: 111621725 | Vijay Vasudevan | vrv@google.com | 1,452,201,962,000 |
cb91829d58a368dd87a00389de91846ba0131608 | Add PartialTensorShape C++ class.
Wraps the TensorShapeProto, but supports a dimension value of -1 as meaning
an "unknown" value. Contains compatibility and merging capabilities similar
to the Python version of TensorShape.
Future changes will add:
* a bool flag to the TensorShape proto that signifies an unknown rank.
* serialization/deserialization of Python TensorShape None values to -1 in the Proto.
* A new Op attr type "partial_shape" that deserializes to PartialTensorShape.
Change: 111622281 | Eugene Brevdo | ebrevdo@gmail.com | 1,452,201,972,000 |
ca47376b3cf19ee395e0409061982266f60e0708 | Make convolutional.py easier to run on GPUs with less RAM.
A few cosmetic fixes. Test error was already 0.7%, not 0.8%.
https://github.com/tensorflow/tensorflow/issues/609
Change: 111624237 | Vincent Vanhoucke | vanhoucke@google.com | 1,452,201,986,000 |
4c4de46cf21d98daeebb897b11efba380d294a0f | Update init_ops to also support initialization of float64 variables.
Change: 111625948 | Yutaka Leon | yutaka.leon@gmail.com | 1,452,201,995,000 |
3cdd5be40751ae04962e37bb581eb1f1ad9ceb4a | Fix a typo
Change: 111628808 | Geoffrey Irving | geoffreyi@google.com | 1,452,202,004,000 |
339db86e2b70a1961e877c59f3a33fa015d5a7a8 | Update generated Op docs.
Change: 111631697 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,202,013,000 |
d38fecedf54d405270377a096f58413101068792 | Further improved the performance of the contrast adjustment code by optimizing the broadcast of scalars
Change: 111631992 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,202,021,000 |
ea2aac69da10fed4acac18dc291790433d9af0ac | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,452,203,413,000 |
02dff6d0d838397860b6ff5256413b54da482996 | Fix bug where attrs with values that are the empty list
were not being properly set via the Python API.
Change: 111635679 | Josh Levenberg | josh11b@tensorflow.org | 1,452,220,674,000 |
f51dea43db9b8d4adaeeadc71eb2f2f48280bb52 | Fix bug in assertProtoEquals / ProtoEq:
The test code skipped testing of the values of key/value pairs in protos.
Change: 111642612 | Eugene Brevdo | ebrevdo@gmail.com | 1,452,220,681,000 |
006d9dc2ee414123762898eac902fa1f86a0eaf9 | Generate docs.
Change: 111646399 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,688,000 |
66c21d60d4becef8c72162015c66492ba975495a | Finish a TODO where we need to support N*T and type list during
graph-to-func conversion.
Change: 111648019 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,696,000 |
b380b27fa090a43efdc7c47aa209c32810586b1a | Adds a name to x-mean computed in tf.moments. This allows the client to extract this tensor from the graph. However the client should not rely on this, and have a fallback plan.
Change: 111648961 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,704,000 |
827163e960e8cb86d3dc3f70434c22713ac9f41c | Change to Relu gradient computation, to allow it to use the output of the Relu instead of its input. The reason is memory consumption: usually the Relu is followed by a Conv layer, which will keep its inputs until the backward pass. The Relu, which produces these inputs, can use them for its own backprop, and doing so will be more memory efficient than having Relu keep its inputs around.
Change: 111649040 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,712,000 |
2b672c4a2f6aeaea8457fd4941f48f5a9e80d283 | Add gradients for DynamicPartition
Change: 111650709 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,733,000 |
20abe178e8b8e8a207f4a052675f0a21f04d5042 | Change: 111653888 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,741,000 |
efd316bf14aaee5ebe85b06d028c05a4628c51e6 | Sort tags on image dashboard
Change: 111655430 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,758,000 |
107853df74593ee0d88e9634b4fc21431d9404a7 | Add string and quantized datatypes to restore operation.
Also add test cases for missing datatypes in save_op_test and restore_op_test.
Change: 111655908 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,767,000 |
ef51295af88a5856d72caa7b8ba6be236b5cc865 | Autogenerated Change: Release TensorBoard at TAG: 8
Change: 111655910 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,220,775,000 |
be5bc79019ee8cf9f8a92f25d7fd71d6acce9aee | TensorFlow: Add more graphdef validation, this time against OpDefs.
The Graph construction code assumes that the GraphDef's NodeDefs match
the OpDef specifications. When creating NodeDefs using NodeDef
builder, we call ValidateNodeDef() function in node_def_util, but when
loading an arbitrary protobuf, we do not. This changes the code to do
some important sanity checking during Session::Create/Extend.
Bumps the version of graph_def to only do this for new graphs.
Also fixes placeholder's shape attr to really be optional, which
it wasn't.
Fix one test that was using an invalid graph.
Change: 111657920 | Vijay Vasudevan | vrv@google.com | 1,452,220,784,000 |
550b62e2b48b4eb31fb13c3d7925eda38d16a522 | Change: 111658449 | Vijay Vasudevan | vrv@google.com | 1,452,220,792,000 |
96fac8a587003c8d2663b812ac3280e9b1911c10 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,452,221,606,000 |
eef2aaa3f7931cc3c0e7bbc6757386cfbcf5e3ef | Fixed saver relative paths for `latest_checkpoint`
This would be cleaner if we made all paths listed in the "latest"
file relative to the its directory, allowing the removal of the
added `os.path.isabs` checks.
That would make the `os.join` in `saver.latest_checkpoint` much less
surprising.
But at least this way, there is no effect on currently working code.
Fixes #571
Change-Id: I47d8536b9b2ed3dcc193d6e6b7f4573a4e22c9b3 | MarkDaoust | daoust.mj@gmail.com | 1,447,957,595,000 |
52e49ac28fb7d3071ac493a2c03f3c0a531d56df | Merge pull request #573 from MarkDaoust/relative-checkpoints | Vijay Vasudevan | vrv@google.com | 1,452,232,579,000 |
72bf16f937c827050ee71ea4cfa7a25e44303246 | Added support for unsorted top-k output, which makes it much more efficient.
This is especially true if k=1 or k is large.
Change: 111680805 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,266,760,000 |
d3e609c7b5816fe88b45897039f754b311ddb253 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,267,718,000 |
a17bb21e76d1c0335feaa77dd2d0639346bdea12 | ci_build: improve user handling inside the container
* create use home with .bazelrc
* create and use user based on uid instead of user name
* add bazel-ci_build-cache to .gitignore
* add optional extra ci build parameters | Jan Prach | jendap@google.com | 1,452,222,983,000 |
05e44812fc9aef96f35a2b297877da69b6e6260b | Merge pull request #66 from google/neural-translation
Fixes #9: Neural translation | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,273,854,000 |
768e4503d9be6c3d945344d8af8dd540d0145131 | Merge pull request #727 from jendap/ci_build
ci_build: improve user handling inside the container | Vijay Vasudevan | vrv@google.com | 1,452,284,158,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.