hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
cbcaf4e85afb0f9eacbb5d2aadb12407786e0e2d | Add support for hinge loss (SVMs) to the Stohastic Dual Coordinate Ascent (SDCA) optimizer.
Change: 116582004 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,386,385,000 |
4547f52379d013385591aee9225921d170786d57 | Add a TODO to Tensor::SummarizeValue
Change: 116583469 | Geoffrey Irving | geoffreyi@google.com | 1,457,387,193,000 |
86dca4f74e32a83c0fb06643497fc848cff21e6d | Add comment about SparseMatMul safety
Change: 116584285 | Geoffrey Irving | geoffreyi@google.com | 1,457,387,660,000 |
972998efe608be7a9c9fad1ba166d52766037d8a | fix docstring for zero_fraction
Change: 116586466 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,388,737,000 |
cc62d992bfaa41744f91ddf00e0c835d375f4f5c | Make RNN calls require a known input depth (via shape inference).
This is a necessary change to ensure that we can remove input_size from
the RNNCell objects.
Change: 116586469 | Eugene Brevdo | ebrevdo@gmail.com | 1,457,388,739,000 |
d82cc200e027552d46f1a0ff193f9cd48992c3f0 | TensorFlow: select op: add support for empty tensors: select
can propagate empty tensors.
Change: 116588842 | Vijay Vasudevan | vrv@google.com | 1,457,390,047,000 |
13024c58fa87a87db23f6869a0c33598bb5c027d | Changed the cuda_crosstool_condition to check for a define of using_cuda.
Checking for a specific crosstool_top directory doesn't work when TensorFlow
is a sub-module for a different project.
Change: 116592676 | Manjunath Kudlur | keveman@gmail.com | 1,457,392,148,000 |
34880c4ca1a83b443b44dc415b74deb217c42257 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,457,393,642,000 |
d94b4795d1fd8d13794c3600d240e3603214a235 | ci_build: support tensorflow as submodule | Jan Prach | jendap@google.com | 1,455,922,823,000 |
3fc885fb1bf57c38c5b80e74f61e9e1e4b3ca089 | Adding current version print to version guard in __init__ | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,417,576,000 |
4d04bc7d0d3f4f89e9fe14e0550267c8debe5696 | Added scripts to generate documentation | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,417,602,000 |
0b7422e07f98c3348ec0e5efbdcfb7829e300505 | Adding initial documentation | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,418,290,000 |
27edec857f3f86af6653a0d510105344643e3d23 | updated rst to md | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,418,374,000 |
22a3bbe00cb57824fc913202fad5822808171580 | added how-to for documentation generation | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,418,728,000 |
c8944c4cac56c7413f0d5f1b8e0dd32048bd9ec6 | Removed redundant instance of StandardScaler
Changed Pipeline to use the existing instance of StandardScaler instead of creating a new instance when Pipeline is created. | Dustin Dorroh | dustindorroh@gmail.com | 1,457,465,540,000 |
a8165fce8f6938910173734e53b42a77c2fe47c1 | Merge pull request #132 from dustindorroh/patch-1
Removed redundant instance of StandardScaler | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,457,467,330,000 |
8312f3eff85ce927c00f7fe4ec1c2fa26da52746 | Fixes #131: set minimum loss to be last_loss_seen | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,457,467,716,000 |
8904ca68fb53ae5b67086cfd0c508f1d9d86ecef | Merge pull request #1206 from jendap/ci_build-reusable-docker
ci_build: support tensorflow as submodule | Jan Prach | jendap@gmail.com | 1,457,470,921,000 |
848d554a78715391421737088c348c513e3e8c6d | Prototype of an in-process gRPC server for TensorFlow/Python.
Adds support for binding a TensorFlow server to any port, to support
single-process testing.
This interface is a work in progress. In particular, it supports
launching a server, but the support for clean shutdown is incomplete.
Change: 116593644 | Derek Murray | mrry@google.com | 1,457,392,699,000 |
01b60af2d61b6ecd77646856c8528275a9dbd0b0 | Assume errors are rare in OP_REQUIRES*() macros.
Change: 116594819 | Josh Levenberg | josh11b@tensorflow.org | 1,457,393,357,000 |
8d8a0c672aa6deb0ceddcf9dbfe52d854bbe8c21 | Add tests for wrong defaults.
Change: 116595261 | Jianmin Chen | goog.jmchen@gmail.com | 1,457,393,606,000 |
624eafe2fb91ac6ae0fba5cfef0cda6e3a79dcfa | Improve the validity of indexing in the segment reduction kernels.
Change: 116595330 | Josh Levenberg | josh11b@tensorflow.org | 1,457,393,649,000 |
204400d34666d09f92cbc0df93fd0504ad4450b8 | Further improvements for training long sequences. Here is the current performance of rnn.dyanmic_rnn() on various sequence lengths. The last column is the time for each time step. It stayed as a constant until len=600, when GPU/CPU memory swapping kicked in. It increased gradually as we had to swap more out of GPU memo... | Yuan Yu | yuanbyu@google.com | 1,457,394,193,000 |
74e06872da2729737767df8fd5923aeeec13af6e | Add appropriate bounds checking to in_top_k op.
Add type-specialized versions of FastBoundsCheck to handle when it's
called on differing type arguments.
Change: 116596487 | David G. Andersen | dga@google.com | 1,457,394,234,000 |
4e00949e5b7469b66505b37f965f36b66b7a41fc | Minor code cleanups.
Change: 116599820 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,396,069,000 |
16f2a29fda4d9efc9d169e443edbcc326a20e162 | Fix (and cleanup) tests in TensorBoard and simplified build process.
Fixed things:
- The internal testing was not testing lib/js/node-radar and components/tf-categorizer.
- The external testing (gulp test) was not running any tests in lib/js/*, and also components/tf-categorizer.
- Now both systems test everything.... | Dan Smilkov | dsmilkov@gmail.com | 1,457,396,981,000 |
9ba61973849a9ef79104c8295886049634a193b4 | A scalable *single-machine* version of SDCA for datasets that don't fit in memory. The optimizer code is now driven from python.
Change: 116602844 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,397,855,000 |
e6388816e30386182967e149a2e59e68c7a73f26 | Remove the requirement for sequence_length input to dynamic_rnn.
Change: 116605455 | Eugene Brevdo | ebrevdo@gmail.com | 1,457,399,479,000 |
c602fd2531cff9366a68a2e7a1d40b13401226e4 | TensorFlow: remove fast_cpp_protos and oversize_protos options from
bazelrc template, since they don't actually do anything.
Change: 116606748 | Vijay Vasudevan | vrv@google.com | 1,457,400,404,000 |
af60815919304f22732fa5f8b13d7c5161f8a41b | Upgraded to a new version of Eigen that adds the ability to pad using values
other than 0 and significantly speeds up a number of computations on GPUs.
Change: 116607765 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,401,186,000 |
19fd2b13725bf43c8e8236696f80b89612a5e879 | Add license headers to server_lib.h server_lib.cc
Change: 116609503 | Vijay Vasudevan | vrv@google.com | 1,457,402,596,000 |
52f49d7cdfde0fff2f52e069e1c588df6c3b9ee9 | TensorFlow: create a CPU device that is only created when GPUs are also
potentially linked into the binary. This makes sure that the :core_cpu
target will never have any GPU code linked in, which was confusing
and weird.
Change: 116618884 | Vijay Vasudevan | vrv@google.com | 1,457,413,035,000 |
cd2d52070f2bb113e77558663f15d55224725c48 | Remove some unnecessary py_api_versions from BUILD files.
Change: 116619279 | Vijay Vasudevan | vrv@google.com | 1,457,413,441,000 |
e98bd22a6c322c7c1f3444e8e0dccff5b358a043 | Correct minor typo in tutorial comment.
Change: 116622681 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,417,663,000 |
439a269fa19459a42c816417e295e3a9c22962cc | Fix a typo: test accuracy -> training accuracy.
Change: 116665650 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,459,097,000 |
484a80ce99998b59cf9f606a7c2a9ad1c14ea29a | - addressing github issue #1198 (reverse_sequence only accepts int64)
Change: 116671488 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,462,118,000 |
febfec7256026a38bda30c78c2768b32e64c9a2c | Partially revert a change to zeros_like which broke the Mandelbrot example
An earlier change made zeros_like inefficient if an output dtype different than
the input was desired: it first made zeros of the input dtype and then cast.
In addition to not being efficient, this broke the Mandelbrot example, which
used zeros... | Geoffrey Irving | geoffreyi@google.com | 1,457,464,127,000 |
93766a718b5309343e011c9efbc50096ddb3b67a | TensorFlow: add ignore_existing option to colocate_with to address
issue of nested colocate_with blocks that are incompatible.
Change: 116684797 | Vijay Vasudevan | vrv@google.com | 1,457,468,827,000 |
d6192acccda3c6f2c9cbc3456856e764a82aa100 | Add bounds check in DecodeCsvOp for when a quoted string is not closed.
Change: 116684950 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,468,934,000 |
f0d630445fbe7fa7538ea6acf52f58cbf12bf1c1 | Remove redundant filegroup definitions from tensorflow/core/BUILD and just use filegroups from tensorflow/core/kernels/BUILD for Android targets.
Change: 116688861 | Andrew Harp | andrew.harp@gmail.com | 1,457,471,445,000 |
8e13e68efcdb96862ce32f9b2a138ca8fac78ec6 | Add stop_loop to LooperThread to enable cleanups.
Change: 116691546 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,472,939,000 |
55aec0a33011773240f6696393952c984ca8de16 | Add explicit not-`None` checks for the height and width in `resize_images()`.
This was previously raising a `FutureWarning` when the height and/or width were dynamic.
Change: 116699992 | Derek Murray | mrry@google.com | 1,457,477,285,000 |
e02403c442cc74673b7d24b5e8104d60f6448c5d | Update Udacity assignments past v0.7.1.
Change: 116703851 | Vincent Vanhoucke | vanhoucke@google.com | 1,457,479,320,000 |
6c576f017156cc0bf50e2dd2cd42b2ab8db9bffe | Add bool transpose on GPU
Fixes #542.
Change: 116704226 | Geoffrey Irving | geoffreyi@google.com | 1,457,479,525,000 |
e1ac9aea1ca0a5aac83fa5b0f1e0929e412e44bd | Use __host__ __device__ attributes during host and device-side CUDA compilation.
Change: 116708519 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,481,893,000 |
d5944fe3725a0e1194bf76c1a29422cbe6c66f29 | Fix breakage from "Assume errors are rare in OP_REQUIRES*() macros"
due to a bad definition of TF_PREDICT_TRUE in macros.h.
Change: 116712226 | Josh Levenberg | josh11b@tensorflow.org | 1,457,483,978,000 |
e48c70ba22dce43fa474c0daeee15019a98a1e3c | Fixes computation of SdcaModel.regularized_loss()
Change: 116714987 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,485,632,000 |
7022c01ac7282eed90db6b2c52195819264bc146 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,457,488,453,000 |
d7a98a5d34bce37010655c61df0353be26648274 | Changes to support tensorflow.github.io/skflow website | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,499,082,000 |
16c9e8dca473f4fb1344647f9d2456aaced19720 | Merge branch 'master' of github.com:google/skflow | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,500,823,000 |
c12b21ca8fb6f38b89fe769176079d8530c8683b | In regular DataFeeder for classification convert y into int64 to not raise numpy warnings on itemset | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,508,961,000 |
78c5ac2894b9a28924211d717e9153edcd5ab160 | Update image_ops.py
Fix Error using tf.image.random._ : 'numpy.ndarray' object has no attribute 'get_shape' #1399 by fix flip_left_to_right to call convert_to_tensor fisrt | Wenjian Huang | nextrush@163.com | 1,457,416,747,000 |
b4b2575ec4bf7e6da2686505f61b5f16cb9273ab | Merge pull request #1428 from fayeshine/fayeshine-patch-for-issue-1399. | Geoffrey Irving | geoffreyi@google.com | 1,457,547,772,000 |
a9a488de5754936a7e67245033f0d6b5fe7b4c88 | Fix #130: Added explicit save/restore to VocabularyProcessor | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,547,980,000 |
2ecd0e448b13964c542a8db23a82e666b519e2f8 | Adding skflow to tensorflow/contrib | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,566,236,000 |
4aeebbba274ca1aa6b54a0aa07075366a5513d41 | Merge branch 'skflow-master' | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,566,266,000 |
5614d3af41fe9bff6b77f0d9b4fbf3723a37d231 | Added skflow to tf.contrib | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,566,309,000 |
fe0181753cf886e14655de7456c3141a01ea3fb7 | Delete old . files | Illia Polosukhin | ilblackdragon@gmail.com | 1,457,566,646,000 |
53ab80e6eefbb06e9bc9d231d5c7e56cee0558d5 | TensorFlow: For errors that are returned without NodeDefs or Ops
associated with them, make sure we return the correct underlying error
code when available.
Change: 116719327 | Vijay Vasudevan | vrv@google.com | 1,457,488,826,000 |
a04bfaed3aa8522f023f91700bb8a33f587ea6ee | Add device_lib swig wrapper which allows accessing the list of local devices.
Change: 116767947 | Eugene Brevdo | ebrevdo@gmail.com | 1,457,542,060,000 |
29144058526b0e5a40f56629dca5f085177f72b6 | Speed up bicubic image resizer op by >2x and make it more readable. Add benchmark.
The main improvements are from hoisting the calls to round() out of the inner loop. Doing the same for the branching associated with Bound() helped as well.
Benchmark Base (ns) New (ns) Improvement
-----------... | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,542,603,000 |
b5c642fb2213651ed4dd03f6e87c3544bebd778f | Upgraded eigen to make it possible to compile a binary that takes advantage of
both avx instructions and cuda to run as fast as possible.
Change: 116775924 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,547,249,000 |
658145f5f18bf1b1741aebeeba245d92bc970f40 | Fixed a couple of compilation warnings
Change: 116776093 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,547,336,000 |
c907e0b3d30c774caee04d71577a9468ed9ab242 | RemoveNewDefaultAttrsFromGraphDef shouldn't touch _underscore attrs
Change: 116782506 | Geoffrey Irving | geoffreyi@google.com | 1,457,550,721,000 |
458e60b1f876e94fa99b056078e3fe4723c9079a | Add tensorflow/core/kernels/cwise_op_rsqrt.cc to android_extended_ops groups
This fix the error: No OpKernel was registered to support Op "Rsqrt" while running model on Android device.
Change: 116783406 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,551,215,000 |
a5923d823e088a9723e445cce9248d5fc59f1b30 | Allow StreamExecutor commands to return status types other than the TensorFlow status type.
Change: 116793254 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,556,707,000 |
fc2ed7dfcc654303a593988cedc05d7ec2b57049 | Fix the crash when applying SparseFtrl update where inner_dimension > 0
Has verified this fix by real job.
Change: 116794658 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,557,534,000 |
0c4be88cc8723e3f07bb8e4b5b918ec1ad1638b2 | Use 32bit indexing to evaluate tiling operations. This improves the performance
on the inception model by about 10% on my gtx titan black.
Change: 116799880 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,560,545,000 |
11f4d83b064fd53b0d4063b3c40212e11710b051 | DecodeRaw should use int64 for sizes, not int
Change: 116802454 | Geoffrey Irving | geoffreyi@google.com | 1,457,561,920,000 |
21be30413d9432e140a234873fa1111d3aa695dd | Replace uses of RE2 with Scanner, for sources that are used on mobile.
This CL also adds the Scanner class to do simple scans over strings, to mimic
regexp behavior like [a-zA-Z][a-zA-Z0-9]* with:
Scanner scan(s);
scan.One(Scanner::LETTER);
scan.Any(Scanner::LETTER_DIGIT);
bool matched = scan.GetResult();
Change:... | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,562,616,000 |
29e75c16ebeb0d9765e9f98a550e2c5b756d9f7b | Adding earlier error handling and better error reporting
of excessively large images to jpeg_mem.
Change: 116809470 | David G. Andersen | dga@google.com | 1,457,565,658,000 |
d21baffdf3d1ef7ea9d36f6b4c6499efc8ba165f | Remove the C++ "grpc_server_lib_test".
This test has been flaky, and the functionality is better tested by
the Python server_lib_test.
Change: 116811917 | Derek Murray | mrry@google.com | 1,457,567,098,000 |
c14ffa94b539c133f4626d7899ea945372a3f901 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,457,581,227,000 |
7de8d0f98ea68ed4959048e14ed1e91c7f4bb2f2 | Adding lbeta, the Logarithm of the multidimensional beta function.
Change: 116815103 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,568,946,000 |
6fbb353d530304ed2eda272a5a968be53dbff676 | Move Android benchmarking state variables to global scope so they can be reset when app is restarted.
Change: 116823789 | Andrew Harp | andrew.harp@gmail.com | 1,457,574,693,000 |
016b25521d0d3462579334f68f34e4052e217bb0 | Fix the FutureWarning in `tf.image.resize_images()` for real this time.
Any of the new and old heights could be `None` (i.e. unknown at graph
construction time) so we must check that they are all non-`None`
before using them in a comparison.
Change: 116824069 | Derek Murray | mrry@google.com | 1,457,574,928,000 |
1a4ff3757aa4b46ad9a13d9070430a3b7ae657e7 | Avoid shuffling, and the use of random generator within the sdca optimizer.
Change: 116824486 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,575,307,000 |
5618915db4a5125632bdabcb2af34ea2b3f46fa2 | Allow restoring Variables whose validate_shape is set to False
Change: 116825422 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,575,955,000 |
49e58a539df6573f926d4147adb586d9f2501444 | Implement gradients wrt sparse values for sparse_tensor_dense_matmul().
Change: 116826347 | Zongheng Yang | zongheng.y@gmail.com | 1,457,576,826,000 |
9f0b11170d6e88c7afb3b11f6165d21144b0e59b | Add MirrorPad op. This op is a variety of Pad op implementing reflect and
symmetric modes of Numpy pad.
Change: 116828726 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,579,403,000 |
c258bca8b08eb1a51e02c4ac1900daf1603850a0 | Use 32bit indexing to pad tensors. This more than doubles the performance of
the corresponding TensorFlow operation.
Change: 116831272 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,582,693,000 |
7706eca44476e8217b61b05d8de70d1ef3e99a5c | Upgraded to a newer version of Eigen that fixes a compilation error on Android
Change: 116831720 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,583,136,000 |
58a72af2bb7b1b51c92ab26edea114c33a91c9ff | Switching to O(N) duplicate counting using unordered_set.
Change: 116831946 | David G. Andersen | dga@google.com | 1,457,583,429,000 |
bb3674caeb8e21e217ffe6446e816608990297bd | Rollback of "Add MirrorPad op. This op is a variety of Pad op implementing reflect and
symmetric modes of Numpy pad."
Change: 116836742 | Vijay Vasudevan | vrv@google.com | 1,457,589,625,000 |
e4237a099f31b99672aaa3d7788f35492cefd420 | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,457,592,658,000 |
236b1c4f2e6de458cc9eecf605b23ca40f839257 | Add missing include that broke the OSS build on Mac.
Change: 116841577 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,596,946,000 |
b145840f1616a0bf17f3369be5379df2a70e0ebb | Fix series node grouping for first item.
Change: 116873663 | James Wexler | james_wexler@yahoo.com | 1,457,628,842,000 |
fdf649e3f7dd3e93506e01b947324f691f2584b9 | Surface control_flow_ops.{foldl, foldr, map_fn} to public.
Change: 116874606 | Yuan Yu | yuanbyu@google.com | 1,457,629,483,000 |
4d9ab3765363298bdcec49f6d617fdfd7708914c | Fix missing spaces in SparseReorder's error messages
Change: 116875393 | Geoffrey Irving | geoffreyi@google.com | 1,457,630,009,000 |
ce076a7743a8bb987bdf73e334d2e474e13af9c6 | Use 32bit indexing to perform average pooling. This improves the performance
of the inception model by almost 10% on my gtx titan black.
Change: 116879126 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,457,632,549,000 |
f699ee7239101947c8d6b5556d43b017259996ed | Move "using namespace tensorflow" declaration in tensorflow_jni.cc so that int64 global variable will be covered.
Change: 116880050 | Andrew Harp | andrew.harp@gmail.com | 1,457,633,069,000 |
0398f4ebe1be0a74a60f9e06035fbc1fdb4444fb | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,457,633,452,000 |
a730b5b2700e7364c6506c65d0ce946864a002a8 | Fixed broken link in "Links"
In "Links" paragraph there is a broken link to tensorflow.org. Fixed it by changing from tensorflow.org to
https://www.tensorflow.org/ | papelita1234 | cachanzheng@gmail.com | 1,457,637,334,000 |
49f4d198789e7ef7bd0d63d7945116014b9c8a33 | ci_build: fixes
* do not require $USER variable
* do not contact github in print_build_info.sh
- there is a easier and faster way to get the clone url | Jan Prach | jendap@google.com | 1,457,341,785,000 |
13ea3ca91ba5aecab6f21acc14b9cb6a9afa8630 | Merge pull request #1447 from jendap/ci_build-fixes. | Martin Wicke | wicke@google.com | 1,457,637,536,000 |
a4b9b51f85d0ac3be317848569b1fdbf530f58f3 | TensorFlow: add more helpful debugging when this check triggers on bad
graphs.
Change: 116880730 | Vijay Vasudevan | vrv@google.com | 1,457,633,416,000 |
0787712b842859c97c13fb7737de8a1bd5d212c7 | Eliminating int conversion warnings from image summary
by limiting the size of images and batches that it will
process (to conform with limits in what jpeg will handle, and
likely png).
Change: 116883534 | David G. Andersen | dga@google.com | 1,457,634,826,000 |
399b80a62e6d64abece8bdb18db2de9dc7cc6e0c | On android, don't preserve doc strings for registered ops. In a binary that
doesn't filter op registrations, this saves >185k. Also, this may save a few
cycles during startup for mobile (untested), since the doc string won't be
parsed.
This introduces use of a TF_LEAN_BINARY macro that we can use to control other
such... | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,637,793,000 |
954876a2392993d1ef294f4f4ea363ddb84c5135 | Allowing contrib kernels to use the kernel bounds checking functions as well.
Change: 116894264 | David G. Andersen | dga@google.com | 1,457,640,421,000 |
3618e3fdde626f16f67bd48dd98bbe0d701c4693 | Change OpDefBuilder to use safe_strto64 for parsing numbers, instead
of doing it itself (incorrectly on mac).
Change: 116897392 | A. Unique TensorFlower | nobody@tensorflow.org | 1,457,642,280,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.