aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
* TensorCoding: Always use meta_ when parsing the RawMessage,Gravatar A. Unique TensorFlower2016-09-17
| | | | | | even in the case of a non-CPU resident value, because meta_ may later be used to test for a dead Tensor. Change: 133500231
* SWIG simplification: Move the FeedVector conversion into C++.Gravatar Derek Murray2016-09-17
| | | | | | | This change moves the %typemap(in) FeedVector into C++ code in tf_session_helper.cc, which now implements the conversion from a string-to-array dictionary into the arguments to `TF_Run()`. Change: 133494142
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-17
| | | | Change: 133492514
* Two tiny TF documentation fixes:Gravatar A. Unique TensorFlower2016-09-17
| | | | | | 1) For softmax: "Computes log softmax activations." -> "Computes softmax activations." 2) For log_softmax: Adds a missing log operation in the documented formula. Change: 133492074
* Remove redundant typedef of NodeAndOutput.Gravatar Derek Murray2016-09-17
| | | | | | | It was already defined in the header. Some compilers are less inclined to give you the benefit of the doubt when there's a type with the same name declared in `tensorflow` and `tensorflow::{anonymous namespace}`. Change: 133487238
* Add a str_util::ConsumeSuffix() utility function.Gravatar A. Unique TensorFlower2016-09-17
| | | | Change: 133476675
* Change for internal compatibility.Gravatar A. Unique TensorFlower2016-09-17
|
* Make highlightedPoints private on scatterWebGL, now the only protected fieldsGravatar Charles Nicholson2016-09-17
| | | | | | | | in scatterWebGL are the methods required to control a scatterWebGLPointsCanvasLabels. This allows scatterWebGLPointsCanvasLabels to become an aggregate class, which will pave the way for general WebGL-enabled visualizers to coexist and render. Change: 133473665
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-17
| | | | Change: 133466188
* Faster transform operation.Gravatar A. Unique TensorFlower2016-09-17
| | | | Change: 133465495
* Fixed a bug in which a server was trying to start even when master was the ↵Gravatar A. Unique TensorFlower2016-09-17
| | | | | | empty string (and therefore a direct session was needed); the result was an exception. Change: 133464309
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133459813
* Update optimize_loss doc string with details of optimizer usage.Gravatar Illia Polosukhin2016-09-16
| | | | Change: 133459446
* graph_constructor: Cosmetic changes.Gravatar Asim Shankar2016-09-16
| | | | | | | | | | | My intent is to be able to re-use most of GraphConstructor to implement the equivalent of python's import_graph_def() in C++. Breaking that intent down into a series of smaller changes that morph GraphConstructor into a form that will work for that, as well as continue to serve the purpose it does today. This particular change should not change behavior in any way. Change: 133456212
* graph_constructor: Remove dead code.Gravatar Asim Shankar2016-09-16
| | | | | | | | | | | Optimizations such as constant-folding and common-subexpression-elimination are applied directly on a Graph* (for example, by GraphOptimizer in tensorflow/core/common_runtime/graph_optimizer.h). I could not find any usage of the GraphConstructorOptions removed in this change. Would like to keep this file focused on conversion from GraphDef to Graph and leave operations on the in-memory representation elsewhere. Change: 133447199
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133443409
* Adaptative SDCAGravatar A. Unique TensorFlower2016-09-16
| | | | | | - add a proabbility sampling function for SDCA (AdaSDCA+) - add LossDerivative and SmoothnessConstant for losses Change: 133443399
* Expose number of partitions to tf.learn SVM Estimator for multi-machineGravatar A. Unique TensorFlower2016-09-16
| | | | | training of SVMs. Change: 133441267
* Fix RandomShuffleQueue random seed when shared_name is provided.Gravatar A. Unique TensorFlower2016-09-16
| | | | | Not providing both graph seed and op seed must use a random seed value according to set_random_seed(). Current behavior chooses a fixed seed value when op seed is not provided, and also ignores the graph seed. Change: 133440862
* Remove unused dependency on numpy.i.Gravatar Derek Murray2016-09-16
| | | | Change: 133439235
* Implement a function to get frequency for android armv7aGravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133434099
* Python error reporting: Copy primary error to bottom of error output.Gravatar Vijay Vasudevan2016-09-16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Fixes https://twitter.com/karpathy/status/765681445637533696 Example new output for a CPU-only binary that executes: ``` import tensorflow as tf s = tf.Session() with tf.device("/gpu:0"): s.run(tf.constant(1.0) + tf.constant(2.0)) ``` is: Traceback (most recent call last): File "baz.py", line 5, in <module> s.run([c]) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 717, in run run_metadata_ptr) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 855, in _run raise RuntimeError('The Session graph is empty. Add operations to the ' RuntimeError: The Session graph is empty. Add operations to the graph before calling run(). (tf)vrv@vijay-linux:~/vrv.tensorflow/tf$ python baz.py Traceback (most recent call last): File "baz.py", line 4, in <module> s.run(tf.constant(1.0) + tf.constant(2.0)) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 717, in run run_metadata_ptr) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 915, in _run feed_dict_string, options, run_metadata) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 965, in _do_run target_list, options, run_metadata) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 985, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors.InvalidArgumentError: Cannot assign a device to node 'add': Could not satisfy explicit device specification '/device:GPU:0' because no devices matching that specification are registered in this process; available devices: /job:localhost/replica:0/task:0/cpu:0 [[Node: add = Add[T=DT_FLOAT, _device="/device:GPU:0"](Const, Const_1)]] Caused by op u'add', defined at: File "baz.py", line 4, in <module> s.run(tf.constant(1.0) + tf.constant(2.0)) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/ops/math_ops.py", line 748, in binary_op_wrapper return func(x, y, name=name) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_math_ops.py", line 71, in add result = _op_def_lib.apply_op("Add", x=x, y=y, name=name) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 749, in apply_op op_def=op_def) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2372, in create_op original_op=self._default_original_op, op_def=op_def) File "REDACTED/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1298, in __init__ self._traceback = _extract_stack() InvalidArgumentError (see above for traceback): Cannot assign a device to node 'add': Could not satisfy explicit device specification '/device:GPU:0' because no devices matching that specification are registered in this process; available devices: /job:localhost/replica:0/task:0/cpu:0 [[Node: add = Add[T=DT_FLOAT, _device="/device:GPU:0"](Const, Const_1)]] The main error now shows up at the end, which is where people are likely to find it more easily. Change: 133429936
* Shard dnn_test to reduce timeout probability.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133426995
* Remove unused variable from fragment shader. Delete some redundant commentsGravatar Charles Nicholson2016-09-16
| | | | | from private methods. Fix an exception when the page is loaded while the mouse is on the GL canvas. Change: 133424898
* Remove write graph calls in SavedModel tests.Gravatar Sukriti Ramesh2016-09-16
| | | | Change: 133424282
* Move tensorflow/core/lib/io/match.{h,cc} to "lib" rather than "lib_internal"Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133423962
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133423449
* Added contract tests to wide-n-deep estimators.Gravatar Mustafa Ispir2016-09-16
| | | | Change: 133422226
* De-flake learn tests.Gravatar A. Unique TensorFlower2016-09-16
| | | | | | | | | - Removed score assertions from nonlinear_test. These will never be stable until we can make TF deterministic, and even if we could they wouldn't be particularly useful tests for an inherently stochastic system. - Removed saver_test. It wasn't testing anything useful. - Removed early_stopping_test in favor of a better unit test for ValidationMonitor. Change: 133420080
* Always color and highlight the sprites at the beginning of ↵Gravatar Charles Nicholson2016-09-16
| | | | | | | | | scatterWebGLPointsCanvasLabels.onRender. Simplify the rendering logic in preparation for reworking selections. Fix bug where colors weren't being correctly reset when the hover / selection set changed. Change: 133419209
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133418348
* Update TransformedDistribution example comment to separately pass ↵Gravatar A. Unique TensorFlower2016-09-16
| | | | | | distribution class and parameters instead of preconstructed Distribution object. Change: 133417454
* Correct the bias introduced by default EMA initialization in score functionGravatar A. Unique TensorFlower2016-09-16
| | | | | mean baseline. Change: 133414673
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133413914
* sampled_softmax_loss eventually calls out to a log-uniform sampler. This ↵Gravatar A. Unique TensorFlower2016-09-16
| | | | | | means that the order of the classes is important and this should be mentioned in the primary documentation. Change: 133410446
* Expand tests for LinearRegressor before refactoring.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133410307
* Stop using static initializer in cpu utils, just removed initializing methodGravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133406514
* Neither scatterWebGL nor scatterWebGLPointsCanvasLayers cache the colorSelectorGravatar Charles Nicholson2016-09-16
| | | | | | | | | or highlightStroke anymore, they only use them as parameters. Split scatterWebGLPointsCanvasLayers.colorSprites up into colorSprites and highlightSprites. Create two materials for scatterWebGLPointsCanvasLayers, one for the picking render pass and one for the color render pass. This simplifies the logic for setting up each render pass, and is clearer, since each material represents a full render state. Now only uniforms are changed between render passes, which is canonical and expected. Change: 133405947
* Fix dtype handling in tf.contrib.metrics.confusion_matrix()Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133403718
* Fix index out of bounds bug in GetNameFromURI()Gravatar Jonathan Hseu2016-09-16
| | | | Change: 133402965
* Include what you use (SWIG edition): strings.i in cpp_shape_inference.iGravatar Derek Murray2016-09-16
| | | | | | Enables this SWIG file to be built standalone, and removes some dead typemaps that fire unhelpfully if you attempt to do so. Change: 133399977
* lintGravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133399706
* Re-enable rnn_test which was accidentally turned off as part of another change.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133399050
* C++ Gradients: Adds gradient functions (and tests) for component-wise unary ops.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133394757
* Update generated Python Op docs.Gravatar A. Unique TensorFlower2016-09-16
| | | | Change: 133391887
* Convert the labels passed to the softmax cross entropy function as needed toGravatar Benoit Steiner2016-09-16
| | | | | make them compatible with the type of the input features Change: 133391271
* Introduces new tf.contrib.metrics for streaming covariance and pearson ↵Gravatar A. Unique TensorFlower2016-09-16
| | | | | | product-moment correlation. Change: 133388403
* Add the interface in steam executor to call cuDNN batch normalization functions.Gravatar Yao Zhang2016-09-15
| | | | Change: 133345765
* Add set of benchmarks for sampling from tf.contrib.distributions.MixtureGravatar Eugene Brevdo2016-09-15
| | | | | | | | * Two benchmarks: one for MVNDiag (lightweight per-distribution construction) and one for MVNFull (complex per-distribution construction). * Minor tweaks to benchmark platform code to add logging and return benchmark statistics. Change: 133341455
* CleanupGravatar A. Unique TensorFlower2016-09-15
| | | | Change: 133341191
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-10 02:27:39 +0000