TensorFlow v1.0.0-rc2 發布,一個表達機器學習算法的接口

TensorFlow 是一個表達機器學習算法的接口，并且是執行算法的實現框架。使用 TensorFlow 表示的計算可以在眾多異構的系統上方便地移植，從移動設別如手機或者平板電腦到成千的GPU計算集群上都可以執行。該系統靈活，可以被用來表示很多的算法包括，深度神經網絡的訓練和推斷算法，也已經被用作科研和應用機器學習系統在若干的計算機科學領域或者其他領域中，例如語言識別、計算機視覺、機器人、信息檢索、自然語言理解、地理信息抽取和計算藥物發現。

更新日志

• XLA (experimental): initial release of XLA, a domain-specific compiler for TensorFlow graphs, that targets CPUs and GPUs.
• TensorFlow Debugger (tfdbg): command-line interface and API.
• New python 3 docker images added.
• Made pip packages pypi compliant. TensorFlow can now be installed by pip install tensorflowcommand.
• Several python API calls have been changed to resemble NumPy more closely.
• New (experimental) Java API.
• Android: new person detection + tracking demo implementing "Scalable Object Detection using Deep Neural Networks" (with additional YOLO object detector support)
• Android: new camera-based image stylization demo based on "A Learned Representation For Artistic Style"
• To help you upgrade your existing TensorFlow Python code to match the API changes below, we have prepared a conversion script.
• TensorFlow/models have been moved to a separate github repository.
• Division and modulus operators (/, //, %) now match Python (flooring) semantics. This applies totf.div and tf.mod as well. To obtain forced integer truncation based behaviors you can usetf.truncatediv and tf.truncatemod.
• tf.divide() is now the recommended division function. tf.div() will remain, but its semantics do not respond to Python 3 or from future mechanisms.
• tf.reverse() now takes indices of axes to be reversed. E.g. tf.reverse(a, [True, False, True]) must now be written as tf.reverse(a, [0, 2]). tf.reverse_v2() will remain until 1.0 final.
• tf.mul, tf.sub and tf.neg are deprecated in favor of tf.multiply, tf.subtract andtf.negative.
• tf.pack and tf.unpack are deprecated in favor of tf.stack and tf.unstack.
• TensorArray.pack and TensorArray.unpack are getting deprecated in favor of TensorArray.stackand TensorArray.unstack.
• The following Python functions have had their arguments changed to use axis when referring to specific dimensions. We have kept the old keyword arguments for compatibility currently, but we will be removing them well before the final 1.0.
  • tf.argmax: dimension becomes axis
  • tf.argmin: dimension becomes axis
  • tf.count_nonzero: reduction_indices becomes axis
  • tf.expand_dims: dim becomes axis
  • tf.reduce_all: reduction_indices becomes axis
  • tf.reduce_any: reduction_indices becomes axis
  • tf.reduce_join: reduction_indices becomes axis
  • tf.reduce_logsumexp: reduction_indices becomes axis
  • tf.reduce_max: reduction_indices becomes axis
  • tf.reduce_mean: reduction_indices becomes axis
  • tf.reduce_min: reduction_indices becomes axis
  • tf.reduce_prod: reduction_indices becomes axis
  • tf.reduce_sum: reduction_indices becomes axis
  • tf.reverse_sequence: batch_dim becomes batch_axis, seq_dim becomes seq_axis
  • tf.sparse_concat: concat_dim becomes axis
  • tf.sparse_reduce_sum: reduction_axes becomes axis
  • tf.sparse_reduce_sum_sparse: reduction_axes becomes axis
  • tf.sparse_split: split_dim becomes axis
• tf.listdiff has been renamed to tf.setdiff1d to match NumPy naming.
• tf.inv has been renamed to be tf.reciprocal (component-wise reciprocal) to avoid confusion withnp.inv which is matrix inversion
• tf.round now uses banker's rounding (round to even) semantics to match NumPy.
• tf.split now takes arguments in a reversed order and with different keywords. In particular, we now match NumPy order as tf.split(value, num_or_size_splits, axis).
• tf.sparse_split now takes arguments in reversed order and with different keywords. In particular we now match NumPy order as tf.sparse_split(sp_input, num_split, axis). NOTE: we have temporarily made tf.sparse_split require keyword arguments.
• tf.concat now takes arguments in reversed order and with different keywords. In particular we now match NumPy order as tf.concat(values, axis, name).
• tf.image.decode_jpeg by default uses the faster DCT method, sacrificing a little fidelity for improved speed. One can revert to the old behavior by specifying the attribute dct_method='INTEGER_ACCURATE'.
• tf.complex_abs has been removed from the Python interface. tf.abs supports complex tensors and should be used instead.
• Template.var_scope property renamed to .variable_scope
• SyncReplicasOptimizer is removed and SyncReplicasOptimizerV2 renamed to SyncReplicasOptimizer.
• tf.zeros_initializer() and tf.ones_initializer() now return a callable that must be called with initializer arguments, in your code replace tf.zeros_initializer with tf.zeros_initializer().
• SparseTensor.shape has been renamed to SparseTensor.dense_shape. Same forSparseTensorValue.shape.
• Replace tf.scalar_summary, tf.histogram_summary, tf.audio_summary, tf.image_summary with tf.summary.scalar, tf.summary.histogram, tf.summary.audio, tf.summary.image, respectively. The new summary ops take name rather than tag as their first argument, meaning summary ops now respect TensorFlow name scopes.
• Replace tf.train.SummaryWriter and tf.train.SummaryWriterCache with tf.summary.FileWriter and tf.summary.FileWriterCache.
• Removes RegisterShape from public API. Use C++ shape function registration instead.
• Deprecated _ref dtypes from the python API.
• In the C++ API (in tensorflow/cc), Input, Output, etc. have moved from the tensorflow::ops namespace to tensorflow.
• Change arg order for {softmax,sparse_softmax,sigmoid}_cross_entropy_with_logits to be (labels, predictions), and force use of named args.
• New op: parallel_stack.
• Introducing common tf io compression options constants for RecordReader/RecordWriter.
• Add sparse_column_with_vocabulary_file, to specify a feature column that transform string features to IDs, where the mapping is defined by a vocabulary file.
• Added index_to_string_table which returns a lookup table that maps indices to strings.
• Add string_to_index_table, which returns a lookup table that matches strings to indices.
• Add a ParallelForWithWorkerId function.
• Add string_to_index_table, which returns a lookup table that matches strings to indices.
• Support restore session from checkpoint files in v2 in contrib/session_bundle.
• Added a tf.contrib.image.rotate function for arbitrary angles.
• Added tf.contrib.framework.filter_variables as a convenience function to filter lists of variables based on regular expressions.
• make_template() takes an optional custom_getter_ param.
• Added comment about how existing directories are handled by recursive_create_dir.
• Added an op for QR factorizations.
• Divides and mods in Python API now use flooring (Python) semantics.
• Android: pre-built libs are now built nightly.
• Android: cmake/gradle build for TensorFlow Inference library under contrib/android/cmake
• Android: Much more robust Session initialization code.
• Android: TF stats now exposed directly in demo and log when debug mode is active
• Android: new/better README.md documentation
• saved_model is available as tf.saved_model.
• Empty op is now stateful.
• Improve speed of scatter_update on the cpu for ASSIGN operations.
• Change reduce_join to treat reduction_indices in the same way as other reduce_ ops.
• Move TensorForestEstimator to contrib/tensor_forest.
• Enable compiler optimizations by default and allow configuration in configure.
• tf.divide now honors the name field.
• Make metrics weight broadcasting more strict.
• Add new queue-like StagingArea and new ops: stage and unstage.

下載

• Source code (zip)
• Source code (tar.gz)

本站原創，轉載時保留以下信息：
本文轉自：深度開源（open-open.com）
原文地址：http://www.baiduhome.net/news/view/41c75dc0