hash stringlengths 40 40 | msg stringlengths 1 131k | author stringlengths 1 33 | email stringlengths 0 57 | date int64 1,447B 1,698B |
|---|---|---|---|---|
969a8575569e7be468a24bbf4bc23f0b86618920 | * dask extractors are consistent with other io extractors | terrytangyuan | terrytangyuan@gmail.com | 1,452,737,001,000 |
0c08c1a9885169809210b40f372d05a9fbee4219 | Merge pull request #73 from terrytangyuan/patch-1
Temporarily omit some io stuff from coverage | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,738,852,000 |
5a5b785eea9041df1aa32cf9ecbdd42c4544623e | Merge pull request #76 from terrytangyuan/dask
Fixed dask_io.py | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,452,740,644,000 |
be1d7243e5623d2c643b4ec34ffa472d8bdd9dbf | * Fix lint issues | terrytangyuan | terrytangyuan@gmail.com | 1,452,741,455,000 |
00214b6bdeb2d9243af84727d7279572d111a84f | Modified the boston.py example so that the standard scaler is fitted over the training set, not the whole dataset, and then applied to the test set | Nicolas Fauchereau | Nicolas.Fauchereau@gmail.com | 1,452,741,650,000 |
61bf812e041b3b6bc815c2d67ca51c5e888a3aff | fix error in coordinator documentation | Carl Vondrick | vondrick@mit.edu | 1,452,751,972,000 |
f9f26e04929ab20653b2b6ce86fcad71f906b193 | Fix Scatter*.png figures
Change: 112109184 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,785,373,000 |
2f5cdba2becd61ff68034a59185d572d2b06327d | Make Android demo default to not saving preview image to improve performance.
Change: 112111470 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,785,383,000 |
861f8f01334d20e998eae9a759c8f5a1e07721ca | Move responsibility for constructing/destructing complex objects from
a helper class in tensor.cc into the allocator class. This allows
experimental devices more control over their memory allocation.
Change: 112111713 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,785,391,000 |
916776a1744d78d65e7c4f52df3bcaa8cf466872 | Refer to cuDNN v2 by its proper name, not CUDNN 6.5 V2, as requested by NVIDIA.
Change: 112120651 | Martin Wicke | wicke@google.com | 1,452,785,400,000 |
d45ed5c9387e64be8d35a41bcaa7f752f6f3eba5 | Add instructions to set up development without reinstallation (fixes #413)
Change: 112120905 | Martin Wicke | wicke@google.com | 1,452,785,407,000 |
47e38b0405e9db850227290f46b73f811b2bd53e | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,785,476,000 |
669790ac3a02c32076f02f43d49e47dd6be30a77 | Merge pull request #770 from cvondrick/patch-1
fix error in coordinator documentation | Martin Wicke | martin.wicke@gmail.com | 1,452,789,824,000 |
dc9616b8af01d5d352589cdca68bb7cb91ac2168 | Fix typos in string and comments of python and C++ code in `models` | Dongjoon Hyun | dongjoon@apache.org | 1,452,794,978,000 |
c94feeab3952bc90ebb9ff1cbd14082eb8182c4a | Merge pull request #774 from dongjoon-hyun/fix_typos_in_models. Closes #774 | Vijay Vasudevan | vrv@google.com | 1,452,796,798,000 |
dc379e24663ad3555aebf5bc52d446d740e3c472 | Updated TensorBoard to work with Python 3 | Joao Felipe Santos | joao.eel@gmail.com | 1,452,546,302,000 |
da4497632f9057a11c9ef9139759bc20b0caadfd | Merge pull request #749 from jfsantos/master. Closes #749 | Vijay Vasudevan | vrv@google.com | 1,452,798,877,000 |
32ed124dc90b909d07ef4105dd3a46883f29d289 | Fixed Numpy return value for Python 3
Using PY_MAJOR_VERSION instead of NUMPY_IMPORT_ARRAY_RETVAL
Fixed SWIG interface for Python 3 support
Make SWIG forward preprocessor directives to generated code | Joao Felipe Santos | joao.eel@gmail.com | 1,452,701,150,000 |
60114c44b01de65ed3403c06671aae65e6e6e169 | Merge pull request #764 from jfsantos/import_array. Closes #764 | Vijay Vasudevan | vrv@google.com | 1,452,798,950,000 |
e35e96dcf7e948b08c5a9d358c99bf4e9b735d7d | Merge pull request #607 from bcoppens/master
Fix TensorBoard not displaying any images | Daniel W Mane | danmane@gmail.com | 1,452,799,544,000 |
100108e572dc72722e13ae325d49d59213577906 | Merge pull request #78 from nicolasfauchereau/master
scaler fitted to the training set only, then applied to test set | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,452,801,787,000 |
92ea3d1120cf58bac27eac3189b3c6ff0eba68e4 | Added generate_checkpoint_state() which returns a checkpoint proto
with model_checkpoint_path containing absolute path or path relative
to save_dir.
Fixes #751. | Sherry Moore | sherrym@google.com | 1,452,812,299,000 |
e1b82f7400b23bcf5e4b495c3ce7db28a0adda97 | Merge pull request #777 from sherrym/saver. Closes #777 | Vijay Vasudevan | vrv@google.com | 1,452,819,136,000 |
fc87c709b90eba0b1a2dc5c389d216858ed88a81 | Updated README.md - change google->tensorflow | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,821,356,000 |
401749c030fce6786f09824f9efd83df9da91607 | Merge pull request #77 from tensorflow/dask
Fix lint issues | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,452,822,371,000 |
7aa5010e948ca24898e9d69ecec512ddeda3f761 | Reverse for vocab processor and handling incorrect utf for byte processor reverse | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,824,359,000 |
41a85614e43e4763ddf2b4e014f5e4004a28f7a3 | Reverse class id to class name for CategoricalVocabulary | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,824,402,000 |
6f237554c7649ef90527713b66b1d40a7b2ae7c7 | fix linters | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,824,445,000 |
d069f3712b1ddeb0b43dca6f1dcd634ea7d0682b | Continue working on neural translation model - added real dataset, train/test split, streaming and processing with byte processor and output results | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,824,488,000 |
05a172d5941a67210109b532a43c08c32ace3226 | Change: 112162312 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,794,585,000 |
90644c9db115ee99a2a14ba462d70b822ec0804d | Run gendocs
Change: 112164149 | Geoffrey Irving | geoffreyi@google.com | 1,452,795,612,000 |
11a8da42cc9aa30983be5c5e0569b09625febf42 | Add tf.cond to public API
There was some concern that it used weird stuff like switch in the
implementation, but it's widely used and the weird stuff can remain *not* part
of the public API even if this is.
Change: 112166819 | Geoffrey Irving | geoffreyi@google.com | 1,452,797,176,000 |
63f9b3e3a587a0a097248a94f27c92b706e3137b | Make tensorflow/core:test public and expose status_test_util.h.
Change: 112167664 | Josh Levenberg | josh11b@tensorflow.org | 1,452,797,623,000 |
31343cb8db60dee45869737dc6271c17668efae0 | TensorFlow: add __pow__ python overload operator.
Change: 112175414 | Vijay Vasudevan | vrv@google.com | 1,452,801,712,000 |
c7258dfbf2f97714596676972bf76880c4db6253 | Move GPU specific code out of generic DMA codepaths.
The stream() and gpu_device_info() calls are not necessary for simple DMAs to/from a device. Moving this code into the GPU-GPU case allows the other DMA codepaths to be used for non-StreamExecutor devices which provide copy methods via their DeviceContext.
Change: 112204599 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,819,254,000 |
7c94dbed882e3f406cea28f46c3c3d8f08f9bcca | Remove an unnecessary warning.
Change: 112210734 | Xiaoqiang Zheng | zhengxq@google.com | 1,452,824,340,000 |
1497b9841bb68bde8bf57ff967c73541bce0c1ae | + Basic test for dask io | terrytangyuan | terrytangyuan@gmail.com | 1,452,827,160,000 |
b378657b8c50f49f9622281e698c62d5538b9397 | Fixes #79: Added custom dropout op, that adds probability tensor to the collection DROPOUTS and zeros it on prediction | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,827,879,000 |
fb6a16bb2cb89c8eb21afed565d27cc275a37c67 | + Tests for dask.dataframe and dask.series | terrytangyuan | terrytangyuan@gmail.com | 1,452,828,013,000 |
3bb833bed156b7f293cea284ceb9a91cea85c7cb | + Minor fixes in dask_io | terrytangyuan | terrytangyuan@gmail.com | 1,452,828,770,000 |
4db219ed59c4be40eda955414307b38e66cb18b6 | - Include dask_io.py in coverage test | terrytangyuan | terrytangyuan@gmail.com | 1,452,828,824,000 |
fd5180833f07dad2908283d3b8c4c5ab3b6fb39f | Merge pull request #82 from tensorflow/dask
dask_io test coverage | Yuan (Terry) Tang | terrytangyuan@users.noreply.github.com | 1,452,830,090,000 |
7b5fef17806e2138ef16e8399ef5b1078d404dbe | Merge commit for internal changes | Vijay Vasudevan | vrv@google.com | 1,452,833,979,000 |
8797e1d4d9f2bdd722c7bc4109f815f25c4ddad6 | Unspecified message. | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,835,503,000 |
aabf28c02a4dff593e5e4b156052adb9b81a70c7 | Test for dropout probability be a tensor | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,838,473,000 |
76ec94160b349b9b006b6fe3704d6be94b9a9efc | Merge pull request #81 from tensorflow/dropout-fix
Fixes #79: Zeros dropouts on prediction | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,838,768,000 |
ac8514f6ec959c727862e7ac1004f2e1fd088420 | Added Mnist example that showcases skflow for tutorial from https://www.tensorflow.org/versions/master/tutorials/mnist/pros/index.html#deep-mnist-for-experts | Illia Polosukhin | ilblackdragon@gmail.com | 1,452,847,207,000 |
87380f76c5f85fa6f0353a27d4f1690b72a79e82 | Fixes. | makseq-ubnt | makseq@gmail.com | 1,452,864,093,000 |
1397271823786b921b6b67b65fdf1936d90276b3 | Fixes for pylint. | makseq-ubnt | makseq@gmail.com | 1,452,871,968,000 |
42d705408beb0d7a45798fa22df5684296716a38 | Unspecified message. | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,875,336,000 |
b76648b7b29c278c38039457172ba5515d94c5b3 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,875,424,000 |
2aec209d1f79d571480401abe77e45742fdb7405 | Use augmented assignment statements and fix typos in mnist/word2vec python code. | Dongjoon Hyun | dongjoon@apache.org | 1,452,835,148,000 |
7ddf1ebfdd120409c19bd6727c18a3f283ada42c | Merge pull request #779 from dongjoon-hyun/use_augmented_assignment_and_fix_typos. Closes #779 | Vijay Vasudevan | vrv@google.com | 1,452,878,321,000 |
409c68eb2531f450680233c437b219237b0609e2 | Fix the split-on-underscore and show-data-download-links options.
The regressions were introduced in cl/110462065
Also fix the tf-tensorboard/demo/index.html file (minor)
Change: 112265591 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,884,581,000 |
b926d941a3d07476df43e070937333fd4af82df6 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,890,466,000 |
621820d41adedc6dc80f37b30c556a5cbe5c3601 | Fix typos in a testcase name and comments of core/common_runtime module. | Dongjoon Hyun | dongjoon@apache.org | 1,452,882,929,000 |
d78f1dfcb5cb1cbd09eacc7257a41c43a27d124d | Merge pull request #787 from dongjoon-hyun/fix_typos_in_common_runtime. Closes #787 | Vijay Vasudevan | vrv@google.com | 1,452,895,434,000 |
f79ed11b282da667b12aeac4e0104a6637254347 | Add instruction to make CUDNN libraries readable. | Frank Chu | fpmc@google.com | 1,451,951,884,000 |
b92e154ff2921d7cb0222c36fad00f1ebdd21965 | Fix typos. | Frank Chu | fpmc@google.com | 1,451,984,115,000 |
a1a3c034c291cca49eeb6ac9710e535db9a1cd9e | Added an option to the attention decoder to initialize the attention
masks from the initial state. This is useful when we wish to resume
decoding from an externally stored state and attention tensor (e.g.,
during beam searching).
Change: 112277385 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,893,082,000 |
3019126ce06d1d8c5269964d279702c42b36efa7 | Removes static shape checks from certain image ops.
The following ops are affected:
* tf.image.flip_left_right()
* tf.image.flip_up_down()
* tf.image.per_image_whitening()
* tf.image.random_flip_left_right()
* tf.image.random_flip_up_down()
* tf.image.transpose_image()
This change allows these ops to be used in a setting where each image
may have a different shape (such as a JPEG-decoding pipeline). When
static shape information is available, the same checks are performed.
To improve the produced shape information, this change also makes the
shape function for `tf.reverse()` more precise when dealing with an
input with partially-known shape.
Change: 112277661 | Derek Murray | mrry@google.com | 1,452,893,329,000 |
0a43371c76c2cbf47a93186acececc576b71c06b | Change: 112277705 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,893,365,000 |
c07a2536e2b8c763401de60a27b1b908d6be299c | Modify LooperThread to wait the specified delay period before the first
invocation of run_loop().
Change: 112279306 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,894,412,000 |
adf57cf86b82cd0a1f0c9e9bef8fb9837c8f343b | Fix broken exception message in the shape function for conv2d.
Change: 112281623 | Derek Murray | mrry@google.com | 1,452,895,535,000 |
6460b5c0ae716d5700ff006aed98dab9ddaaccdb | Current sorting criterion in the PTB example is not deterministic for words with the same counts, so the vocabulary is different for each run. This is a problem when one wants to modify this code to add save/restore functionalities as the model is then random for the words which have the same count.
Change: 112286934 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,898,784,000 |
19c56143f3286e46e2bedaef6257092b380c8856 | Improve resources page: call out all subpages in index.md.
Change: 112288822 | Martin Wicke | wicke@google.com | 1,452,899,928,000 |
d3d3e33796435e30d285a39ab34513f5b0b2cafd | Adds {batch_}matrix_solve and {batch_}triangular_matrix_solve ops.
Change: 112289410 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,900,361,000 |
64c9d61fa97ba09cf9f52c0b5213d4d9103f0c8b | Handles the case where a kernel does not output anything yet it
promises to.
Change: 112290467 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,901,133,000 |
8cdbaccc15f723e2ccb02fc4ee51b57983503aad | Merge pull request #789 from fpmchu/docs
Add instructions for chmod for linux cuDNN installation, and documentation typo fixes. | Martin Wicke | martin.wicke@gmail.com | 1,452,905,698,000 |
2a2d6c33a44ffc47c18dda8816b48b7432fc8c72 | * Get around with bug in dask | terrytangyuan | terrytangyuan@gmail.com | 1,452,905,932,000 |
08dab0a13502db70a30ee4ba463107e2ea7ad4c4 | * Fixed DaskDataFeeder | terrytangyuan | terrytangyuan@gmail.com | 1,452,907,130,000 |
2a128fcaa6eb2771428f300c728db4db647b5927 | Fix missing dependency declarations error #469 #772 | Yi-Lin Juang | frankyjuang@gmail.com | 1,452,917,152,000 |
139a3afa829755d10dc97715dfa98edd18fea9be | Merge pull request #788 from frankyjuang/master. Closes #788 | Vijay Vasudevan | vrv@google.com | 1,452,910,229,000 |
70b39040fbf19bde61aa41828a083ced66aafaba | Change: 112295685 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,905,024,000 |
bd3af957185f2835c638a045c80d6c11b7f71df1 | Slight efficiency improvements for setting up binary op kernels.
Change: 112296712 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,905,713,000 |
2302df7c73dac7bcc3dc3a69baa00498819fb23d | Generalize top_k to any rank >= 1 and Tensor k
Previously, top_k only worked for matrices, and required the k value to be
known at Graph construction time. This CL lifts both restrictions. Since
changing an attr to an input is a backwards incompatible change, TopK still has
an attr and there is a new TopKV2 that takes an input.
Since the GraphDef versioning mechanism is in the middle of a redesign, I
haven't used OP_DEPRECATE on the old TopK version just yet. Instead, the
Python wrapper invokes the new one only if the input is a Tensor. This is
temporary.
Change: 112297740 | Geoffrey Irving | geoffreyi@google.com | 1,452,906,369,000 |
0cc8848a52ac4131db0b3884bc52c06f0039969c | Upgrade to the latest version of eigen which improves the performance of
matrix-vector products and provides various fixes for CUDA.
Change: 112298067 | Benoit Steiner | benoit.steiner.goog@gmail.com | 1,452,906,660,000 |
e75bd059e5f4c76e2e0fd7d5220962e99cd1d960 | Add the UniqueWithCounts op. This is the same as the Unique op but has an additional output that returns the number of duplicate values.
Change: 112301474 | Andrew Dai | adai@google.com | 1,452,910,028,000 |
966ab186b6ee30b5276269ec2924efa018b61f6c | TensorFlow: add python3 requirements to classify_image.py
Change: 112302259 | Vijay Vasudevan | vrv@google.com | 1,452,911,030,000 |
67b82da322acd410b00d08bc05997ecbdb88717f | + Test for DaskDataFeeder | terrytangyuan | terrytangyuan@gmail.com | 1,452,911,773,000 |
7a434bfe304df7797ca2441364fdf3ac0be59442 | * Working dask_data_feeder test | terrytangyuan | terrytangyuan@gmail.com | 1,452,913,521,000 |
ad0a2eda417bb667782d776328da22bd57589197 | * Fixed lint | terrytangyuan | terrytangyuan@gmail.com | 1,452,913,580,000 |
ba13fd9a7468a045baeee9265a325228d5c5df02 | * Finished test for dask_data_feeder | terrytangyuan | terrytangyuan@gmail.com | 1,452,915,209,000 |
e847a69598c3aae08e1a35de0d3a8793f717ca63 | + Turn on coverage test for data feeder | terrytangyuan | terrytangyuan@gmail.com | 1,452,915,604,000 |
87c56a216bfc6b50d85ab5b10917b9cba508f0f9 | Merge branch 'dask' | terrytangyuan | terrytangyuan@gmail.com | 1,452,916,903,000 |
d3d67cfbac0778ba9b27776cb2ccc5b66ad0820a | + DaskDataFeeder support multi-dim | terrytangyuan | terrytangyuan@gmail.com | 1,452,922,071,000 |
cbd042dd505778da038d79f69451da1fba729296 | Change TensorBoard BOWER files to visibility Public
Change: 112303694 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,913,261,000 |
6e1c71daf60947d8b1b3d5da229fe66bf688ab92 | Autogenerated Change: Release TensorBoard at TAG: 9
Change: 112304018 | A. Unique TensorFlower | nobody@tensorflow.org | 1,452,913,697,000 |
7edc313a15195b495bf23bfdc630ff870f0d6b72 | Change: 112309029 | Manjunath Kudlur | keveman@gmail.com | 1,452,923,812,000 |
0e7d21f04c345d473558146415d1bfaa32b439c7 | Merge commit for internal changes | Manjunath Kudlur | keveman@gmail.com | 1,452,923,995,000 |
4d5d90ec8587fa6cb071945b8d6e2c540b894905 | - minor changes in test | terrytangyuan | terrytangyuan@gmail.com | 1,452,954,315,000 |
230568c53a63032e6bc74d88418899da41616bd2 | * minor refactoring | terrytangyuan | terrytangyuan@gmail.com | 1,452,955,047,000 |
798c02a86e6a094a1167b48102428a9988ac8ab5 | + Integrate DaskDataFeeder with estimators | terrytangyuan | terrytangyuan@gmail.com | 1,452,957,611,000 |
a4f6b3d51617be69beef46523ffb3812239e2a91 | Update index.md
fix missing comma | Vijay Vasudevan | vrv@google.com | 1,452,972,598,000 |
16ae059c7680208ed96db3515f614d9b35a2b25a | Fix typos in python/training module. | Dongjoon Hyun | dongjoon@apache.org | 1,452,837,239,000 |
5ab60166d386246126fe7ea611ee4c24d7262c78 | * Fixed column name issue and sample_fraction | terrytangyuan | terrytangyuan@gmail.com | 1,452,997,371,000 |
b485dbd68b4a2870ff662d4993cb0f85759641c6 | + Fixed repeated colnames when concat two dfs with default numeric colnames | terrytangyuan | terrytangyuan@gmail.com | 1,453,005,461,000 |
51cd15779c95c8acf31802599d498f1a598f0f53 | * Fixed lint | terrytangyuan | terrytangyuan@gmail.com | 1,453,005,672,000 |
f2bd0fc399606d14b55f3f7d732d013f32b33dd5 | Merge pull request #793 from dongjoon-hyun/fix_typos_in_python_training. Closes #793 | Vijay Vasudevan | vrv@google.com | 1,453,009,700,000 |
0428af32a36623b5a7d093aaa0198b4d407e081b | Print the correct host information instead of `localhost` alias.
If a user specify a host ip by using `--host` option,
any connections to `http://localhost:6006` are refused. | Dongjoon Hyun | dongjoon@apache.org | 1,452,904,552,000 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.