├── .gitignore ├── LICENSE ├── README.md ├── doc ├── en │ ├── Makefile │ ├── make.bat │ ├── requirements.txt │ └── source │ │ ├── _static │ │ ├── css │ │ │ └── pytorch_theme.css │ │ └── img │ │ │ ├── dynamic_graph.gif │ │ │ ├── pytorch-logo-dark.png │ │ │ ├── pytorch-logo-dark.svg │ │ │ ├── pytorch-logo-flame.png │ │ │ ├── pytorch-logo-flame.svg │ │ │ └── tensor_illustration.png │ │ ├── _templates │ │ └── layout.html │ │ ├── autograd.rst │ │ ├── conf.py │ │ ├── cuda.rst │ │ ├── data.rst │ │ ├── distributed.rst │ │ ├── distributions.rst │ │ ├── ffi.rst │ │ ├── index.rst │ │ ├── legacy.rst │ │ ├── model_zoo.rst │ │ ├── multiprocessing.rst │ │ ├── nn.rst │ │ ├── notes │ │ ├── autograd.rst │ │ ├── broadcasting.rst │ │ ├── cuda.rst │ │ ├── extending.rst │ │ ├── multiprocessing.rst │ │ └── serialization.rst │ │ ├── onnx.rst │ │ ├── optim.rst │ │ ├── sparse.rst │ │ ├── storage.rst │ │ ├── tensors.rst │ │ ├── torch.rst │ │ └── torchvision │ │ ├── datasets.rst │ │ ├── index.rst │ │ ├── models.rst │ │ ├── transforms.rst │ │ └── utils.rst └── zh │ ├── Makefile │ ├── make.bat │ ├── requirements.txt │ └── source │ ├── _static │ ├── css │ │ └── pytorch_theme.css │ └── img │ │ ├── dynamic_graph.gif │ │ ├── pytorch-logo-dark.png │ │ ├── pytorch-logo-dark.svg │ │ ├── pytorch-logo-flame.png │ │ ├── pytorch-logo-flame.svg │ │ └── tensor_illustration.png │ ├── _templates │ └── layout.html │ ├── apachecn-learning-group.rst │ ├── autograd.rst │ ├── code │ ├── torch │ │ ├── __init__.py │ │ ├── _six.py │ │ ├── _storage_docs.py │ │ ├── _tensor_docs.py │ │ ├── _tensor_str.py │ │ ├── _thnn │ │ │ ├── __init__.py │ │ │ └── utils.py │ │ ├── _torch_docs.py │ │ ├── _utils.py │ │ ├── autograd │ │ │ ├── __init__.py │ │ │ ├── _functions │ │ │ │ ├── __init__.py │ │ │ │ ├── basic_ops.py │ │ │ │ ├── blas.py │ │ │ │ ├── compare.py │ │ │ │ ├── initializers.py │ │ │ │ ├── linalg.py │ │ │ │ ├── pointwise.py │ │ │ │ ├── reduce.py │ │ │ │ ├── stochastic.py │ │ │ │ ├── tensor.py │ │ │ │ └── utils.py │ │ │ ├── function.py │ │ │ ├── gradcheck.py │ │ │ ├── profiler.py │ │ │ ├── stochastic_function.py │ │ │ └── variable.py │ │ ├── backends │ │ │ ├── __init__.py │ │ │ └── cudnn │ │ │ │ ├── __init__.py │ │ │ │ └── rnn.py │ │ ├── contrib │ │ │ ├── __init__.py │ │ │ └── _graph_vis.py │ │ ├── cuda │ │ │ ├── __init__.py │ │ │ ├── comm.py │ │ │ ├── error.py │ │ │ ├── nccl.py │ │ │ ├── nvtx.py │ │ │ ├── profiler.py │ │ │ ├── random.py │ │ │ ├── sparse.py │ │ │ └── streams.py │ │ ├── distributed │ │ │ ├── __init__.py │ │ │ └── remote_types.py │ │ ├── distributions.py │ │ ├── for_onnx │ │ │ └── __init__.py │ │ ├── functional.py │ │ ├── jit │ │ │ ├── __init__.py │ │ │ └── passes │ │ │ │ ├── __init__.py │ │ │ │ └── inplace.py │ │ ├── legacy │ │ │ ├── __init__.py │ │ │ ├── nn │ │ │ │ ├── Abs.py │ │ │ │ ├── AbsCriterion.py │ │ │ │ ├── Add.py │ │ │ │ ├── AddConstant.py │ │ │ │ ├── BCECriterion.py │ │ │ │ ├── BatchNormalization.py │ │ │ │ ├── Bilinear.py │ │ │ │ ├── CAddTable.py │ │ │ │ ├── CDivTable.py │ │ │ │ ├── CMul.py │ │ │ │ ├── CMulTable.py │ │ │ │ ├── CSubTable.py │ │ │ │ ├── Clamp.py │ │ │ │ ├── ClassNLLCriterion.py │ │ │ │ ├── ClassSimplexCriterion.py │ │ │ │ ├── Concat.py │ │ │ │ ├── ConcatTable.py │ │ │ │ ├── Container.py │ │ │ │ ├── Contiguous.py │ │ │ │ ├── Copy.py │ │ │ │ ├── Cosine.py │ │ │ │ ├── CosineDistance.py │ │ │ │ ├── CosineEmbeddingCriterion.py │ │ │ │ ├── Criterion.py │ │ │ │ ├── CriterionTable.py │ │ │ │ ├── CrossEntropyCriterion.py │ │ │ │ ├── DepthConcat.py │ │ │ │ ├── DistKLDivCriterion.py │ │ │ │ ├── DotProduct.py │ │ │ │ ├── Dropout.py │ │ │ │ ├── ELU.py │ │ │ │ ├── Euclidean.py │ │ │ │ ├── Exp.py │ │ │ │ ├── FlattenTable.py │ │ │ │ ├── GradientReversal.py │ │ │ │ ├── HardShrink.py │ │ │ │ ├── HardTanh.py │ │ │ │ ├── HingeEmbeddingCriterion.py │ │ │ │ ├── Identity.py │ │ │ │ ├── Index.py │ │ │ │ ├── JoinTable.py │ │ │ │ ├── L1Cost.py │ │ │ │ ├── L1HingeEmbeddingCriterion.py │ │ │ │ ├── L1Penalty.py │ │ │ │ ├── LeakyReLU.py │ │ │ │ ├── Linear.py │ │ │ │ ├── Log.py │ │ │ │ ├── LogSigmoid.py │ │ │ │ ├── LogSoftMax.py │ │ │ │ ├── LookupTable.py │ │ │ │ ├── MM.py │ │ │ │ ├── MSECriterion.py │ │ │ │ ├── MV.py │ │ │ │ ├── MarginCriterion.py │ │ │ │ ├── MarginRankingCriterion.py │ │ │ │ ├── MaskedSelect.py │ │ │ │ ├── Max.py │ │ │ │ ├── Mean.py │ │ │ │ ├── Min.py │ │ │ │ ├── MixtureTable.py │ │ │ │ ├── Module.py │ │ │ │ ├── Mul.py │ │ │ │ ├── MulConstant.py │ │ │ │ ├── MultiCriterion.py │ │ │ │ ├── MultiLabelMarginCriterion.py │ │ │ │ ├── MultiLabelSoftMarginCriterion.py │ │ │ │ ├── MultiMarginCriterion.py │ │ │ │ ├── Narrow.py │ │ │ │ ├── NarrowTable.py │ │ │ │ ├── Normalize.py │ │ │ │ ├── PReLU.py │ │ │ │ ├── Padding.py │ │ │ │ ├── PairwiseDistance.py │ │ │ │ ├── Parallel.py │ │ │ │ ├── ParallelCriterion.py │ │ │ │ ├── ParallelTable.py │ │ │ │ ├── PartialLinear.py │ │ │ │ ├── Power.py │ │ │ │ ├── RReLU.py │ │ │ │ ├── ReLU.py │ │ │ │ ├── ReLU6.py │ │ │ │ ├── Replicate.py │ │ │ │ ├── Reshape.py │ │ │ │ ├── Select.py │ │ │ │ ├── SelectTable.py │ │ │ │ ├── Sequential.py │ │ │ │ ├── Sigmoid.py │ │ │ │ ├── SmoothL1Criterion.py │ │ │ │ ├── SoftMarginCriterion.py │ │ │ │ ├── SoftMax.py │ │ │ │ ├── SoftMin.py │ │ │ │ ├── SoftPlus.py │ │ │ │ ├── SoftShrink.py │ │ │ │ ├── SoftSign.py │ │ │ │ ├── SpatialAdaptiveMaxPooling.py │ │ │ │ ├── SpatialAveragePooling.py │ │ │ │ ├── SpatialBatchNormalization.py │ │ │ │ ├── SpatialClassNLLCriterion.py │ │ │ │ ├── SpatialContrastiveNormalization.py │ │ │ │ ├── SpatialConvolution.py │ │ │ │ ├── SpatialConvolutionLocal.py │ │ │ │ ├── SpatialConvolutionMap.py │ │ │ │ ├── SpatialCrossMapLRN.py │ │ │ │ ├── SpatialDilatedConvolution.py │ │ │ │ ├── SpatialDivisiveNormalization.py │ │ │ │ ├── SpatialDropout.py │ │ │ │ ├── SpatialFractionalMaxPooling.py │ │ │ │ ├── SpatialFullConvolution.py │ │ │ │ ├── SpatialFullConvolutionMap.py │ │ │ │ ├── SpatialLPPooling.py │ │ │ │ ├── SpatialMaxPooling.py │ │ │ │ ├── SpatialMaxUnpooling.py │ │ │ │ ├── SpatialReflectionPadding.py │ │ │ │ ├── SpatialReplicationPadding.py │ │ │ │ ├── SpatialSoftMax.py │ │ │ │ ├── SpatialSubSampling.py │ │ │ │ ├── SpatialSubtractiveNormalization.py │ │ │ │ ├── SpatialUpSamplingNearest.py │ │ │ │ ├── SpatialZeroPadding.py │ │ │ │ ├── SplitTable.py │ │ │ │ ├── Sqrt.py │ │ │ │ ├── Square.py │ │ │ │ ├── Squeeze.py │ │ │ │ ├── Sum.py │ │ │ │ ├── Tanh.py │ │ │ │ ├── TanhShrink.py │ │ │ │ ├── TemporalConvolution.py │ │ │ │ ├── TemporalMaxPooling.py │ │ │ │ ├── TemporalSubSampling.py │ │ │ │ ├── Threshold.py │ │ │ │ ├── Transpose.py │ │ │ │ ├── Unsqueeze.py │ │ │ │ ├── View.py │ │ │ │ ├── VolumetricAveragePooling.py │ │ │ │ ├── VolumetricBatchNormalization.py │ │ │ │ ├── VolumetricConvolution.py │ │ │ │ ├── VolumetricDropout.py │ │ │ │ ├── VolumetricFullConvolution.py │ │ │ │ ├── VolumetricMaxPooling.py │ │ │ │ ├── VolumetricMaxUnpooling.py │ │ │ │ ├── VolumetricReplicationPadding.py │ │ │ │ ├── WeightedEuclidean.py │ │ │ │ ├── WeightedMSECriterion.py │ │ │ │ ├── __init__.py │ │ │ │ └── utils.py │ │ │ └── optim │ │ │ │ ├── __init__.py │ │ │ │ ├── adadelta.py │ │ │ │ ├── adagrad.py │ │ │ │ ├── adam.py │ │ │ │ ├── adamax.py │ │ │ │ ├── asgd.py │ │ │ │ ├── cg.py │ │ │ │ ├── lbfgs.py │ │ │ │ ├── nag.py │ │ │ │ ├── rmsprop.py │ │ │ │ ├── rprop.py │ │ │ │ └── sgd.py │ │ ├── multiprocessing │ │ │ ├── __init__.py │ │ │ ├── pool.py │ │ │ ├── queue.py │ │ │ └── reductions.py │ │ ├── nn │ │ │ ├── __init__.py │ │ │ ├── _functions │ │ │ │ ├── __init__.py │ │ │ │ ├── dropout.py │ │ │ │ ├── linear.py │ │ │ │ ├── loss.py │ │ │ │ ├── padding.py │ │ │ │ ├── rnn.py │ │ │ │ ├── thnn │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── activation.py │ │ │ │ │ ├── auto.py │ │ │ │ │ ├── auto_double_backwards.py │ │ │ │ │ ├── auto_symbolic.py │ │ │ │ │ ├── batchnorm_double_backwards.py │ │ │ │ │ ├── loss.py │ │ │ │ │ ├── normalization.py │ │ │ │ │ ├── pooling.py │ │ │ │ │ ├── rnnFusedPointwise.py │ │ │ │ │ ├── sparse.py │ │ │ │ │ └── upsampling.py │ │ │ │ └── vision.py │ │ │ ├── backends │ │ │ │ ├── __init__.py │ │ │ │ ├── backend.py │ │ │ │ └── thnn.py │ │ │ ├── functional.py │ │ │ ├── init.py │ │ │ ├── modules │ │ │ │ ├── __init__.py │ │ │ │ ├── activation.py │ │ │ │ ├── batchnorm.py │ │ │ │ ├── container.py │ │ │ │ ├── conv.py │ │ │ │ ├── distance.py │ │ │ │ ├── dropout.py │ │ │ │ ├── instancenorm.py │ │ │ │ ├── linear.py │ │ │ │ ├── loss.py │ │ │ │ ├── module.py │ │ │ │ ├── normalization.py │ │ │ │ ├── padding.py │ │ │ │ ├── pixelshuffle.py │ │ │ │ ├── pooling.py │ │ │ │ ├── rnn.py │ │ │ │ ├── sparse.py │ │ │ │ ├── upsampling.py │ │ │ │ └── utils.py │ │ │ ├── parallel │ │ │ │ ├── __init__.py │ │ │ │ ├── _functions.py │ │ │ │ ├── data_parallel.py │ │ │ │ ├── distributed.py │ │ │ │ ├── parallel_apply.py │ │ │ │ ├── replicate.py │ │ │ │ └── scatter_gather.py │ │ │ ├── parameter.py │ │ │ └── utils │ │ │ │ ├── __init__.py │ │ │ │ ├── clip_grad.py │ │ │ │ ├── convert_parameters.py │ │ │ │ ├── rnn.py │ │ │ │ └── weight_norm.py │ │ ├── onnx │ │ │ ├── __init__.py │ │ │ └── symbolic.py │ │ ├── optim │ │ │ ├── __init__.py │ │ │ ├── adadelta.py │ │ │ ├── adagrad.py │ │ │ ├── adam.py │ │ │ ├── adamax.py │ │ │ ├── asgd.py │ │ │ ├── lbfgs.py │ │ │ ├── lr_scheduler.py │ │ │ ├── optimizer.py │ │ │ ├── rmsprop.py │ │ │ ├── rprop.py │ │ │ ├── sgd.py │ │ │ └── sparse_adam.py │ │ ├── random.py │ │ ├── serialization.py │ │ ├── sparse │ │ │ └── __init__.py │ │ ├── storage.py │ │ ├── tensor.py │ │ ├── utils │ │ │ ├── __init__.py │ │ │ ├── backcompat │ │ │ │ └── __init__.py │ │ │ ├── data │ │ │ │ ├── __init__.py │ │ │ │ ├── dataloader.py │ │ │ │ ├── dataset.py │ │ │ │ ├── distributed.py │ │ │ │ └── sampler.py │ │ │ ├── dlpack.py │ │ │ ├── ffi │ │ │ │ └── __init__.py │ │ │ ├── hooks.py │ │ │ ├── model_zoo.py │ │ │ ├── serialization │ │ │ │ ├── __init__.py │ │ │ │ └── read_lua_file.py │ │ │ └── trainer │ │ │ │ ├── __init__.py │ │ │ │ ├── plugins │ │ │ │ ├── __init__.py │ │ │ │ ├── accuracy.py │ │ │ │ ├── logger.py │ │ │ │ ├── loss.py │ │ │ │ ├── monitor.py │ │ │ │ ├── plugin.py │ │ │ │ ├── progress.py │ │ │ │ └── time.py │ │ │ │ └── trainer.py │ │ └── version.py │ └── torchvision │ │ ├── __init__.py │ │ ├── datasets │ │ ├── __init__.py │ │ ├── cifar.py │ │ ├── coco.py │ │ ├── fakedata.py │ │ ├── folder.py │ │ ├── lsun.py │ │ ├── mnist.py │ │ ├── phototour.py │ │ ├── semeion.py │ │ ├── stl10.py │ │ ├── svhn.py │ │ └── utils.py │ │ ├── models │ │ ├── __init__.py │ │ ├── alexnet.py │ │ ├── densenet.py │ │ ├── inception.py │ │ ├── resnet.py │ │ ├── squeezenet.py │ │ └── vgg.py │ │ ├── transforms.py │ │ ├── transforms │ │ ├── __init__.py │ │ ├── functional.py │ │ └── transforms.py │ │ └── utils.py │ ├── conf.py │ ├── cuda.rst │ ├── data.rst │ ├── distributed.rst │ ├── distributions.rst │ ├── ffi.rst │ ├── index.rst │ ├── legacy.rst │ ├── model_zoo.rst │ ├── multiprocessing.rst │ ├── nn.rst │ ├── notes │ ├── autograd.rst │ ├── broadcasting.rst │ ├── cuda.rst │ ├── extending.rst │ ├── multiprocessing.rst │ └── serialization.rst │ ├── onnx.rst │ ├── optim.rst │ ├── project-contributors.rst │ ├── sparse.rst │ ├── storage.rst │ ├── tensors.rst │ ├── torch.rst │ └── torchvision │ ├── datasets.rst │ ├── index.rst │ ├── models.rst │ ├── transforms.rst │ └── utils.rst ├── example └── tutorials │ └── beginner │ └── deep_learning_60min_bltiz.ipynb └── tutorial ├── en ├── .gitignore ├── LICENSE ├── Makefile ├── README.md ├── _static │ ├── css │ │ └── pytorch_theme.css │ └── img │ │ ├── SRResNet.png │ │ ├── Variable.png │ │ ├── cartpole.gif │ │ ├── cat.jpg │ │ ├── cat_224x224.jpg │ │ ├── cat_output1.png │ │ ├── char_rnn_generation.png │ │ ├── cifar10.png │ │ ├── data_parallel.png │ │ ├── distributed │ │ ├── DistPyTorch.jpg │ │ ├── all_gather.pdf │ │ ├── all_gather.png │ │ ├── all_reduce.pdf │ │ ├── all_reduce.png │ │ ├── broadcast.png │ │ ├── gather.png │ │ ├── reduce.png │ │ ├── scatter.png │ │ ├── send_recv.png │ │ └── send_recv_big.png │ │ ├── dynamic_graph.gif │ │ ├── landmarked_face2.png │ │ ├── mnist.png │ │ ├── neural-style │ │ ├── dancing.jpg │ │ ├── neuralstyle.png │ │ └── picasso.jpg │ │ ├── pytorch-logo-dark.png │ │ ├── pytorch-logo-dark.svg │ │ ├── seq-seq-images │ │ ├── attention-decoder-network.dot │ │ ├── attention-decoder-network.png │ │ ├── decoder-network.dot │ │ ├── decoder-network.png │ │ ├── decoder.png │ │ ├── decoder@2x.png │ │ ├── encoder-network.dot │ │ ├── encoder-network.png │ │ ├── seq2seq.png │ │ ├── seq2seq@2x.png │ │ ├── word-encoding.png │ │ └── word-encoding@2x.png │ │ ├── seq2seq_flat.png │ │ ├── stn │ │ ├── FSeq.png │ │ ├── Five.gif │ │ ├── stn-arch.png │ │ └── tr.png │ │ ├── tensor_illustration.png │ │ ├── tensor_illustration_flat.png │ │ ├── thumbnails │ │ ├── babel.jpg │ │ ├── default.png │ │ ├── examples.png │ │ ├── pytorch-logo-flat.png │ │ └── torch-logo.png │ │ └── torch-nn-vs-pytorch-nn.png ├── _templates │ └── layout.html ├── advanced_source │ ├── README.txt │ ├── c_extension.rst │ ├── neural_style_tutorial.py │ ├── numpy_extensions_tutorial.py │ └── super_resolution_with_caffe2.py ├── beginner_source │ ├── README.txt │ ├── blitz │ │ ├── README.txt │ │ ├── autograd_tutorial.py │ │ ├── cifar10_tutorial.py │ │ ├── data_parallel_tutorial.py │ │ ├── neural_networks_tutorial.py │ │ └── tensor_tutorial.py │ ├── data_loading_tutorial.py │ ├── deep_learning_60min_blitz.rst │ ├── deep_learning_nlp_tutorial.rst │ ├── examples_autograd │ │ ├── README.txt │ │ ├── tf_two_layer_net.py │ │ ├── two_layer_net_autograd.py │ │ └── two_layer_net_custom_function.py │ ├── examples_nn │ │ ├── README.txt │ │ ├── dynamic_net.py │ │ ├── two_layer_net_module.py │ │ ├── two_layer_net_nn.py │ │ └── two_layer_net_optim.py │ ├── examples_tensor │ │ ├── README.txt │ │ ├── two_layer_net_numpy.py │ │ └── two_layer_net_tensor.py │ ├── former_torchies │ │ ├── README.txt │ │ ├── autograd_tutorial.py │ │ ├── nn_tutorial.py │ │ ├── parallelism_tutorial.py │ │ └── tensor_tutorial.py │ ├── former_torchies_tutorial.rst │ ├── nlp │ │ ├── README.txt │ │ ├── advanced_tutorial.py │ │ ├── deep_learning_tutorial.py │ │ ├── pytorch_tutorial.py │ │ ├── sequence_models_tutorial.py │ │ └── word_embeddings_tutorial.py │ ├── pytorch_with_examples.rst │ └── transfer_learning_tutorial.py ├── build.sh ├── conf.py ├── custom_directives.py ├── index.rst ├── intermediate_source │ ├── README.txt │ ├── char_rnn_classification_tutorial.py │ ├── char_rnn_generation_tutorial.py │ ├── dist_tuto.rst │ ├── reinforcement_q_learning.py │ ├── seq2seq_translation_tutorial.py │ └── spatial_transformer_tutorial.py └── requirements.txt └── zh ├── .gitignore ├── LICENSE ├── Makefile ├── README.md ├── _static ├── css │ └── pytorch_theme.css └── img │ ├── SRResNet.png │ ├── Variable.png │ ├── cartpole.gif │ ├── cat.jpg │ ├── cat_224x224.jpg │ ├── cat_output1.png │ ├── char_rnn_generation.png │ ├── cifar10.png │ ├── data_parallel.png │ ├── distributed │ ├── DistPyTorch.jpg │ ├── all_gather.pdf │ ├── all_gather.png │ ├── all_reduce.pdf │ ├── all_reduce.png │ ├── broadcast.png │ ├── gather.png │ ├── reduce.png │ ├── scatter.png │ ├── send_recv.png │ └── send_recv_big.png │ ├── dynamic_graph.gif │ ├── landmarked_face2.png │ ├── mnist.png │ ├── neural-style │ ├── dancing.jpg │ ├── neuralstyle.png │ └── picasso.jpg │ ├── pytorch-logo-dark.png │ ├── pytorch-logo-dark.svg │ ├── seq-seq-images │ ├── attention-decoder-network.dot │ ├── attention-decoder-network.png │ ├── decoder-network.dot │ ├── decoder-network.png │ ├── decoder.png │ ├── decoder@2x.png │ ├── encoder-network.dot │ ├── encoder-network.png │ ├── seq2seq.png │ ├── seq2seq@2x.png │ ├── word-encoding.png │ └── word-encoding@2x.png │ ├── seq2seq_flat.png │ ├── stn │ ├── FSeq.png │ ├── Five.gif │ ├── stn-arch.png │ └── tr.png │ ├── tensor_illustration.png │ ├── tensor_illustration_flat.png │ ├── thumbnails │ ├── babel.jpg │ ├── default.png │ ├── examples.png │ ├── pytorch-logo-flat.png │ └── torch-logo.png │ └── torch-nn-vs-pytorch-nn.png ├── _templates └── layout.html ├── advanced_source ├── README.txt ├── c_extension.rst ├── neural_style_tutorial.py ├── numpy_extensions_tutorial.py └── super_resolution_with_caffe2.py ├── apachecn-learning-group.rst ├── beginner_source ├── README.txt ├── blitz │ ├── README.txt │ ├── autograd_tutorial.py │ ├── cifar10_tutorial.py │ ├── data_parallel_tutorial.py │ ├── neural_networks_tutorial.py │ └── tensor_tutorial.py ├── data_loading_tutorial.py ├── deep_learning_60min_blitz.rst ├── deep_learning_nlp_tutorial.rst ├── examples_autograd │ ├── README.txt │ ├── tf_two_layer_net.py │ ├── two_layer_net_autograd.py │ └── two_layer_net_custom_function.py ├── examples_nn │ ├── README.txt │ ├── dynamic_net.py │ ├── two_layer_net_module.py │ ├── two_layer_net_nn.py │ └── two_layer_net_optim.py ├── examples_tensor │ ├── README.txt │ ├── two_layer_net_numpy.py │ └── two_layer_net_tensor.py ├── former_torchies │ ├── README.txt │ ├── autograd_tutorial.py │ ├── nn_tutorial.py │ ├── parallelism_tutorial.py │ └── tensor_tutorial.py ├── former_torchies_tutorial.rst ├── nlp │ ├── README.txt │ ├── advanced_tutorial.py │ ├── deep_learning_tutorial.py │ ├── pytorch_tutorial.py │ ├── sequence_models_tutorial.py │ └── word_embeddings_tutorial.py ├── pytorch_with_examples.rst └── transfer_learning_tutorial.py ├── build.sh ├── conf.py ├── custom_directives.py ├── index.rst ├── intermediate_source ├── README.txt ├── char_rnn_classification_tutorial.py ├── char_rnn_generation_tutorial.py ├── dist_tuto.rst ├── reinforcement_q_learning.py ├── seq2seq_translation_tutorial.py └── spatial_transformer_tutorial.py ├── project-contributors.rst └── requirements.txt /doc/en/Makefile: -------------------------------------------------------------------------------- 1 | # Minimal makefile for Sphinx documentation 2 | # 3 | 4 | # You can set these variables from the command line. 5 | SPHINXOPTS = 6 | SPHINXBUILD = sphinx-build 7 | SPHINXPROJ = PyTorch 8 | SOURCEDIR = source 9 | BUILDDIR = build 10 | 11 | # Put it first so that "make" without argument is like "make help". 12 | help: 13 | @$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) 14 | 15 | docset: html 16 | doc2dash --name $(SPHINXPROJ) --icon $(SOURCEDIR)/_static/img/pytorch-logo-flame.png --enable-js --online-redirect-url http://pytorch.org/docs/ --force $(BUILDDIR)/html/ 17 | 18 | # Manually fix because Zeal doesn't deal well with `icon.png`-only at 2x resolution. 19 | cp $(SPHINXPROJ).docset/icon.png $(SPHINXPROJ).docset/icon@2x.png 20 | convert $(SPHINXPROJ).docset/icon@2x.png -resize 16x16 $(SPHINXPROJ).docset/icon.png 21 | 22 | .PHONY: help Makefile docset 23 | 24 | # Catch-all target: route all unknown targets to Sphinx using the new 25 | # "make mode" option. $(O) is meant as a shortcut for $(SPHINXOPTS). 26 | %: Makefile 27 | @$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) 28 | -------------------------------------------------------------------------------- /doc/en/make.bat: -------------------------------------------------------------------------------- 1 | @ECHO OFF 2 | 3 | pushd %~dp0 4 | 5 | REM Command file for Sphinx documentation 6 | 7 | if "%SPHINXBUILD%" == "" ( 8 | set SPHINXBUILD=sphinx-build 9 | ) 10 | set SOURCEDIR=source 11 | set BUILDDIR=build 12 | set SPHINXPROJ=PyTorch 13 | 14 | if "%1" == "" goto help 15 | 16 | %SPHINXBUILD% >NUL 2>NUL 17 | if errorlevel 9009 ( 18 | echo. 19 | echo.The 'sphinx-build' command was not found. Make sure you have Sphinx 20 | echo.installed, then set the SPHINXBUILD environment variable to point 21 | echo.to the full path of the 'sphinx-build' executable. Alternatively you 22 | echo.may add the Sphinx directory to PATH. 23 | echo. 24 | echo.If you don't have Sphinx installed, grab it from 25 | echo.http://sphinx-doc.org/ 26 | exit /b 1 27 | ) 28 | 29 | %SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% 30 | goto end 31 | 32 | :help 33 | %SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% 34 | 35 | :end 36 | popd 37 | -------------------------------------------------------------------------------- /doc/en/requirements.txt: -------------------------------------------------------------------------------- 1 | sphinx 2 | -e git://github.com/snide/sphinx_rtd_theme.git#egg=sphinx_rtd_theme 3 | -------------------------------------------------------------------------------- /doc/en/source/_static/img/dynamic_graph.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/en/source/_static/img/dynamic_graph.gif -------------------------------------------------------------------------------- /doc/en/source/_static/img/pytorch-logo-dark.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/en/source/_static/img/pytorch-logo-dark.png -------------------------------------------------------------------------------- /doc/en/source/_static/img/pytorch-logo-flame.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/en/source/_static/img/pytorch-logo-flame.png -------------------------------------------------------------------------------- /doc/en/source/_static/img/pytorch-logo-flame.svg: -------------------------------------------------------------------------------- 1 | 2 | image/svg+xml -------------------------------------------------------------------------------- /doc/en/source/_static/img/tensor_illustration.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/en/source/_static/img/tensor_illustration.png -------------------------------------------------------------------------------- /doc/en/source/_templates/layout.html: -------------------------------------------------------------------------------- 1 | {% extends "!layout.html" %} 2 | 3 | {% block footer %} 4 | {{ super() }} 5 | 6 | 13 | 22 | 36 | {% endblock %} -------------------------------------------------------------------------------- /doc/en/source/cuda.rst: -------------------------------------------------------------------------------- 1 | torch.cuda 2 | =================================== 3 | 4 | .. currentmodule:: torch.cuda 5 | 6 | .. automodule:: torch.cuda 7 | :members: 8 | 9 | Random Number Generator 10 | ------------------------- 11 | .. autofunction:: get_rng_state 12 | .. autofunction:: set_rng_state 13 | .. autofunction:: manual_seed 14 | .. autofunction:: manual_seed_all 15 | .. autofunction:: seed 16 | .. autofunction:: seed_all 17 | .. autofunction:: initial_seed 18 | 19 | 20 | Communication collectives 21 | ------------------------- 22 | 23 | .. autofunction:: torch.cuda.comm.broadcast 24 | 25 | .. autofunction:: torch.cuda.comm.reduce_add 26 | 27 | .. autofunction:: torch.cuda.comm.scatter 28 | 29 | .. autofunction:: torch.cuda.comm.gather 30 | 31 | Streams and events 32 | ------------------ 33 | 34 | .. autoclass:: Stream 35 | :members: 36 | 37 | .. autoclass:: Event 38 | :members: 39 | 40 | Memory management 41 | ----------------- 42 | .. autofunction:: empty_cache 43 | 44 | NVIDIA Tools Extension (NVTX) 45 | ----------------------------- 46 | 47 | .. autofunction:: torch.cuda.nvtx.mark 48 | .. autofunction:: torch.cuda.nvtx.range_push 49 | .. autofunction:: torch.cuda.nvtx.range_pop 50 | -------------------------------------------------------------------------------- /doc/en/source/data.rst: -------------------------------------------------------------------------------- 1 | torch.utils.data 2 | =================================== 3 | 4 | .. automodule:: torch.utils.data 5 | .. autoclass:: Dataset 6 | .. autoclass:: TensorDataset 7 | .. autoclass:: ConcatDataset 8 | .. autoclass:: DataLoader 9 | .. autoclass:: torch.utils.data.sampler.Sampler 10 | .. autoclass:: torch.utils.data.sampler.SequentialSampler 11 | .. autoclass:: torch.utils.data.sampler.RandomSampler 12 | .. autoclass:: torch.utils.data.sampler.SubsetRandomSampler 13 | .. autoclass:: torch.utils.data.sampler.WeightedRandomSampler 14 | .. autoclass:: torch.utils.data.distributed.DistributedSampler 15 | -------------------------------------------------------------------------------- /doc/en/source/distributions.rst: -------------------------------------------------------------------------------- 1 | .. role:: hidden 2 | :class: hidden-section 3 | 4 | Probability distributions - torch.distributions 5 | ================================================== 6 | 7 | .. automodule:: torch.distributions 8 | .. currentmodule:: torch.distributions 9 | 10 | :hidden:`Distribution` 11 | ~~~~~~~~~~~~~~~~~~~~~~~ 12 | 13 | .. autoclass:: Distribution 14 | :members: 15 | 16 | :hidden:`Bernoulli` 17 | ~~~~~~~~~~~~~~~~~~~~~~~ 18 | 19 | .. autoclass:: Bernoulli 20 | :members: 21 | 22 | :hidden:`Categorical` 23 | ~~~~~~~~~~~~~~~~~~~~~~~ 24 | 25 | .. autoclass:: Categorical 26 | :members: 27 | 28 | :hidden:`Normal` 29 | ~~~~~~~~~~~~~~~~~~~~~~~ 30 | 31 | .. autoclass:: Normal 32 | :members: 33 | -------------------------------------------------------------------------------- /doc/en/source/ffi.rst: -------------------------------------------------------------------------------- 1 | torch.utils.ffi 2 | =============== 3 | 4 | .. currentmodule:: torch.utils.ffi 5 | .. autofunction:: create_extension 6 | 7 | -------------------------------------------------------------------------------- /doc/en/source/index.rst: -------------------------------------------------------------------------------- 1 | .. PyTorch documentation master file, created by 2 | sphinx-quickstart on Fri Dec 23 13:31:47 2016. 3 | You can adapt this file completely to your liking, but it should at least 4 | contain the root `toctree` directive. 5 | 6 | :github_url: https://github.com/pytorch/pytorch 7 | 8 | PyTorch documentation 9 | =================================== 10 | 11 | PyTorch is an optimized tensor library for deep learning using GPUs and CPUs. 12 | 13 | .. toctree:: 14 | :glob: 15 | :maxdepth: 1 16 | :caption: Notes 17 | 18 | notes/* 19 | 20 | 21 | .. toctree:: 22 | :maxdepth: 1 23 | :caption: Package Reference 24 | 25 | torch 26 | tensors 27 | sparse 28 | storage 29 | nn 30 | optim 31 | torch.autograd 32 | torch.distributions 33 | torch.multiprocessing 34 | torch.distributed 35 | torch.legacy 36 | cuda 37 | ffi 38 | data 39 | model_zoo 40 | onnx 41 | 42 | .. toctree:: 43 | :glob: 44 | :maxdepth: 2 45 | :caption: torchvision Reference 46 | 47 | torchvision/index 48 | 49 | 50 | Indices and tables 51 | ================== 52 | 53 | * :ref:`genindex` 54 | * :ref:`modindex` 55 | -------------------------------------------------------------------------------- /doc/en/source/legacy.rst: -------------------------------------------------------------------------------- 1 | Legacy package - torch.legacy 2 | =================================== 3 | 4 | .. automodule:: torch.legacy 5 | -------------------------------------------------------------------------------- /doc/en/source/model_zoo.rst: -------------------------------------------------------------------------------- 1 | torch.utils.model_zoo 2 | =================================== 3 | 4 | .. automodule:: torch.utils.model_zoo 5 | .. autofunction:: load_url 6 | -------------------------------------------------------------------------------- /doc/en/source/notes/serialization.rst: -------------------------------------------------------------------------------- 1 | 2 | Serialization semantics 3 | ======================= 4 | 5 | Best practices 6 | -------------- 7 | 8 | .. _recommend-saving-models: 9 | 10 | Recommended approach for saving a model 11 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 12 | 13 | There are two main approaches for serializing and restoring a model. 14 | 15 | The first (recommended) saves and loads only the model parameters:: 16 | 17 | torch.save(the_model.state_dict(), PATH) 18 | 19 | Then later:: 20 | 21 | the_model = TheModelClass(*args, **kwargs) 22 | the_model.load_state_dict(torch.load(PATH)) 23 | 24 | The second saves and loads the entire model:: 25 | 26 | torch.save(the_model, PATH) 27 | 28 | Then later:: 29 | 30 | the_model = torch.load(PATH) 31 | 32 | However in this case, the serialized data is bound to the specific classes 33 | and the exact directory structure used, so it can break in various ways when 34 | used in other projects, or after some serious refactors. 35 | -------------------------------------------------------------------------------- /doc/en/source/storage.rst: -------------------------------------------------------------------------------- 1 | torch.Storage 2 | =================================== 3 | 4 | A :class:`torch.Storage` is a contiguous, one-dimensional array of a single 5 | data type. 6 | 7 | Every :class:`torch.Tensor` has a corresponding storage of the same data type. 8 | 9 | .. autoclass:: torch.FloatStorage 10 | :members: 11 | :undoc-members: 12 | :inherited-members: 13 | -------------------------------------------------------------------------------- /doc/en/source/torchvision/index.rst: -------------------------------------------------------------------------------- 1 | torchvision 2 | =========== 3 | 4 | The :mod:`torchvision` package consists of popular datasets, model 5 | architectures, and common image transformations for computer vision. 6 | 7 | .. toctree:: 8 | :maxdepth: 2 9 | :caption: Package Reference 10 | 11 | datasets 12 | models 13 | transforms 14 | utils 15 | 16 | .. automodule:: torchvision 17 | :members: -------------------------------------------------------------------------------- /doc/en/source/torchvision/transforms.rst: -------------------------------------------------------------------------------- 1 | torchvision.transforms 2 | ====================== 3 | 4 | .. currentmodule:: torchvision.transforms 5 | 6 | Transforms are common image transforms. They can be chained together using :class:`Compose` 7 | 8 | .. autoclass:: Compose 9 | 10 | Transforms on PIL Image 11 | ----------------------- 12 | 13 | .. autoclass:: Resize 14 | 15 | .. autoclass:: Scale 16 | 17 | .. autoclass:: CenterCrop 18 | 19 | .. autoclass:: RandomCrop 20 | 21 | .. autoclass:: RandomHorizontalFlip 22 | 23 | .. autoclass:: RandomVerticalFlip 24 | 25 | .. autoclass:: RandomResizedCrop 26 | 27 | .. autoclass:: RandomSizedCrop 28 | 29 | .. autoclass:: Grayscale 30 | 31 | .. autoclass:: RandomGrayscale 32 | 33 | .. autoclass:: FiveCrop 34 | 35 | .. autoclass:: TenCrop 36 | 37 | .. autoclass:: Pad 38 | 39 | .. autoclass:: ColorJitter 40 | 41 | Transforms on torch.\*Tensor 42 | ---------------------------- 43 | 44 | .. autoclass:: Normalize 45 | :members: __call__ 46 | :special-members: 47 | 48 | 49 | Conversion Transforms 50 | --------------------- 51 | 52 | .. autoclass:: ToTensor 53 | :members: __call__ 54 | :special-members: 55 | 56 | .. autoclass:: ToPILImage 57 | :members: __call__ 58 | :special-members: 59 | 60 | Generic Transforms 61 | ------------------ 62 | 63 | .. autoclass:: Lambda -------------------------------------------------------------------------------- /doc/en/source/torchvision/utils.rst: -------------------------------------------------------------------------------- 1 | torchvision.utils 2 | ================= 3 | 4 | .. currentmodule:: torchvision.utils 5 | 6 | .. autofunction:: make_grid 7 | 8 | .. autofunction:: save_image -------------------------------------------------------------------------------- /doc/zh/Makefile: -------------------------------------------------------------------------------- 1 | # Minimal makefile for Sphinx documentation 2 | # 3 | 4 | # You can set these variables from the command line. 5 | SPHINXOPTS = 6 | SPHINXBUILD = sphinx-build 7 | SPHINXPROJ = PyTorch 8 | SOURCEDIR = source 9 | BUILDDIR = build 10 | 11 | # Put it first so that "make" without argument is like "make help". 12 | help: 13 | @$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) 14 | 15 | docset: html 16 | doc2dash --name $(SPHINXPROJ) --icon $(SOURCEDIR)/_static/img/pytorch-logo-flame.png --enable-js --online-redirect-url http://pytorch.org/docs/ --force $(BUILDDIR)/html/ 17 | 18 | # Manually fix because Zeal doesn't deal well with `icon.png`-only at 2x resolution. 19 | cp $(SPHINXPROJ).docset/icon.png $(SPHINXPROJ).docset/icon@2x.png 20 | convert $(SPHINXPROJ).docset/icon@2x.png -resize 16x16 $(SPHINXPROJ).docset/icon.png 21 | 22 | .PHONY: help Makefile docset 23 | 24 | # Catch-all target: route all unknown targets to Sphinx using the new 25 | # "make mode" option. $(O) is meant as a shortcut for $(SPHINXOPTS). 26 | %: Makefile 27 | @$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O) 28 | -------------------------------------------------------------------------------- /doc/zh/make.bat: -------------------------------------------------------------------------------- 1 | @ECHO OFF 2 | 3 | pushd %~dp0 4 | 5 | REM Command file for Sphinx documentation 6 | 7 | if "%SPHINXBUILD%" == "" ( 8 | set SPHINXBUILD=sphinx-build 9 | ) 10 | set SOURCEDIR=source 11 | set BUILDDIR=build 12 | set SPHINXPROJ=PyTorch 13 | 14 | if "%1" == "" goto help 15 | 16 | %SPHINXBUILD% >NUL 2>NUL 17 | if errorlevel 9009 ( 18 | echo. 19 | echo.The 'sphinx-build' command was not found. Make sure you have Sphinx 20 | echo.installed, then set the SPHINXBUILD environment variable to point 21 | echo.to the full path of the 'sphinx-build' executable. Alternatively you 22 | echo.may add the Sphinx directory to PATH. 23 | echo. 24 | echo.If you don't have Sphinx installed, grab it from 25 | echo.http://sphinx-doc.org/ 26 | exit /b 1 27 | ) 28 | 29 | %SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% 30 | goto end 31 | 32 | :help 33 | %SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% 34 | 35 | :end 36 | popd 37 | -------------------------------------------------------------------------------- /doc/zh/requirements.txt: -------------------------------------------------------------------------------- 1 | sphinx 2 | -e git://github.com/snide/sphinx_rtd_theme.git#egg=sphinx_rtd_theme 3 | -------------------------------------------------------------------------------- /doc/zh/source/_static/img/dynamic_graph.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/_static/img/dynamic_graph.gif -------------------------------------------------------------------------------- /doc/zh/source/_static/img/pytorch-logo-dark.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/_static/img/pytorch-logo-dark.png -------------------------------------------------------------------------------- /doc/zh/source/_static/img/pytorch-logo-flame.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/_static/img/pytorch-logo-flame.png -------------------------------------------------------------------------------- /doc/zh/source/_static/img/pytorch-logo-flame.svg: -------------------------------------------------------------------------------- 1 | 2 | image/svg+xml -------------------------------------------------------------------------------- /doc/zh/source/_static/img/tensor_illustration.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/_static/img/tensor_illustration.png -------------------------------------------------------------------------------- /doc/zh/source/_templates/layout.html: -------------------------------------------------------------------------------- 1 | {% extends "!layout.html" %} 2 | 3 | {% block footer %} 4 | {{ super() }} 5 | 6 | 13 | 22 | 36 | {% endblock %} -------------------------------------------------------------------------------- /doc/zh/source/apachecn-learning-group.rst: -------------------------------------------------------------------------------- 1 | 组织学习交流群 2 | ================== 3 | 4 | 机器学习交流群: `629470233 `__ （2000人） 5 | 6 | 大数据交流群: `214293307 `__ （2000人） 7 | 8 | Kaggle 竞赛交流群: `686932392 `__ （2000人） 9 | 10 | 了解我们: ``__ 11 | 12 | 加入组织: ``__ 13 | 14 | 更多学（zhuang）习（bi）交流群请参阅: ``__ -------------------------------------------------------------------------------- /doc/zh/source/autograd.rst: -------------------------------------------------------------------------------- 1 | .. role:: hidden 2 | :class: hidden-section 3 | 4 | Automatic differentiation package - torch.autograd 5 | ================================================== 6 | 7 | .. automodule:: torch.autograd 8 | .. currentmodule:: torch.autograd 9 | 10 | .. autofunction:: backward 11 | 12 | .. autofunction:: grad 13 | 14 | Variable（变量） 15 | ---------------- 16 | 17 | API compatibility 18 | ^^^^^^^^^^^^^^^^^ 19 | 20 | Variable API 几乎与常规 Tensor API 相同(一些会覆盖梯度计算输入的内置方法除外). 21 | 在大多数情况下, 变量量可以安全地替换张量并且代码将保持正常工作. 22 | 因为这个, 我们没有记录变量上的所有操作, 你应该参阅 :class:`torch.Tensor` 文档来查看变量上的所有操作. 23 | 24 | In-place operations on Variables 25 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 26 | 27 | 在 autograd 支持就地操作是一件困难的事情, 在大多数情况下我们不鼓励使用. 28 | Autograd 积极的缓冲区释放和重用使得它非常高效, 而且很少有就地操作实际上大量地降低了内存使用量的情况. 29 | 除非你正在大量的的内存压力下运行, 否则你可能永远不需要使用它们. 30 | 31 | In-place correctness checks 32 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ 33 | 34 | 所有的 :class:`Variable` 跟踪适用于它们的就地操作, 并且如果实现检测到一个变量是否被其中一个函数后台保存, 但是之后它被就地修改了, 会在开始求导时会报出异常. 35 | 这确保了如果你在就地使用函数并没有看到任何错误, 你可以肯定的是计算变量是正确的. 36 | 37 | 38 | .. autoclass:: Variable 39 | :members: 40 | 41 | :hidden:'Function(函数)' 42 | --------------------------- 43 | 44 | .. autoclass:: Function 45 | :members: 46 | 47 | Profiler(分析器) 48 | ------------------ 49 | 50 | Autograd 包含一个分析器, 可以让你检查你的模型在CPU 和 GPU 上不同运算的成本. 51 | 目前实现有两种模式 - 只使用 CPU 的 :class:`~torch.autograd.profiler.profile`. 52 | 和基于 nvprof（注册 CPU 和 GPU 活动）的方式使用 :class:`~torch.autograd.profiler.emit_nvtx`. 53 | 54 | .. autoclass:: torch.autograd.profiler.profile 55 | :members: 56 | 57 | .. autoclass:: torch.autograd.profiler.emit_nvtx 58 | :members: 59 | 60 | .. autofunction:: torch.autograd.profiler.load_nvprof 61 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/_storage_docs.py: -------------------------------------------------------------------------------- 1 | """Adds docstrings to Storage functions""" 2 | 3 | import torch._C 4 | from torch._C import _add_docstr as add_docstr 5 | 6 | 7 | storage_classes = [ 8 | 'DoubleStorageBase', 9 | 'FloatStorageBase', 10 | 'LongStorageBase', 11 | 'IntStorageBase', 12 | 'ShortStorageBase', 13 | 'CharStorageBase', 14 | 'ByteStorageBase', 15 | ] 16 | 17 | 18 | def add_docstr_all(method, docstr): 19 | for cls_name in storage_classes: 20 | cls = getattr(torch._C, cls_name) 21 | try: 22 | add_docstr(getattr(cls, method), docstr) 23 | except AttributeError: 24 | pass 25 | 26 | 27 | add_docstr_all('from_file', 28 | """ 29 | from_file(filename, shared=False, size=0) -> Storage 30 | 31 | 如果 shared 为 True, 那么内存将会在所有进程间共享. 所有的更改都会被写入文件. 如果 shared 为 False, 32 | 那么对于内存的修改, 则不会影响到文件. 33 | 34 | size 是存储中所包含的元素个数. 如果 shared 为 False 则文件必须包含至少 `size * sizeof(Type)` 字节 35 | ( `Type` 是所存储的类型)如果 shared 为 True, 文件会在需要的时候被创建. 36 | 37 | Args: 38 | filename (str): 要映射到的文件名 39 | shared (bool): 是否共享内存 40 | size (int): 存储中包含元素的个数 41 | """) 42 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/autograd/_functions/__init__.py: -------------------------------------------------------------------------------- 1 | from .basic_ops import * 2 | from .tensor import * 3 | from .pointwise import * 4 | from .reduce import * 5 | from .linalg import * 6 | from .blas import * 7 | from .stochastic import * 8 | from .compare import * 9 | from .initializers import * 10 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/autograd/_functions/compare.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | from ..function import Function 4 | from .utils import maybe_unexpand, maybe_unexpand_or_view 5 | 6 | 7 | # TODO: once Cpp-style functions are implemented we can detach a and b 8 | # before calling forward. 9 | class _CompareOp(Function): 10 | 11 | @classmethod 12 | def forward(cls, ctx, a, b): 13 | ctx.a_size = a.size() 14 | ctx.b_tensor = torch.is_tensor(b) 15 | ctx.b_size = b.size() if ctx.b_tensor else None 16 | ctx.input_type = type(a) 17 | mask = getattr(a, cls.fn_name)(b) 18 | ctx.mark_non_differentiable(mask) 19 | return mask 20 | 21 | @staticmethod 22 | def backward(ctx, grad_output): 23 | grad_input = (grad_output * 0).type(ctx.input_type) 24 | return (maybe_unexpand(grad_input, ctx.a_size), 25 | maybe_unexpand_or_view(grad_input, ctx.b_size) if ctx.b_tensor else None) 26 | 27 | 28 | class Eq(_CompareOp): 29 | fn_name = 'eq' 30 | 31 | 32 | class Ne(_CompareOp): 33 | fn_name = 'ne' 34 | 35 | 36 | class Gt(_CompareOp): 37 | fn_name = 'gt' 38 | 39 | 40 | class Ge(_CompareOp): 41 | fn_name = 'ge' 42 | 43 | 44 | class Lt(_CompareOp): 45 | fn_name = 'lt' 46 | 47 | 48 | class Le(_CompareOp): 49 | fn_name = 'le' 50 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/autograd/_functions/initializers.py: -------------------------------------------------------------------------------- 1 | from ..function import InplaceFunction 2 | 3 | 4 | class Zero(InplaceFunction): 5 | 6 | @staticmethod 7 | def forward(ctx, i, inplace=False): 8 | if inplace: 9 | ctx.mark_dirty(i) 10 | result = i.zero_() 11 | else: 12 | result = i.new(1).zero_().expand_as(i) 13 | 14 | ctx.save_for_backward(result) 15 | return result 16 | 17 | @staticmethod 18 | def backward(ctx, grad_output): 19 | result, = ctx.saved_variables 20 | return Variable(result.data.new(1).zero_().expand_as(result)) 21 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/autograd/_functions/stochastic.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from ..function import Function 3 | 4 | 5 | class Categorical(Function): 6 | @staticmethod 7 | def forward(ctx, probs, num_samples, with_replacement): 8 | samples = probs.multinomial(num_samples, with_replacement) 9 | ctx.mark_non_differentiable(samples) 10 | return samples 11 | 12 | @staticmethod 13 | def backward(ctx, grad_output): 14 | return None, None, None 15 | 16 | 17 | class Bernoulli(Function): 18 | @staticmethod 19 | def forward(ctx, probs): 20 | samples = probs.new().resize_as_(probs).bernoulli_(probs) 21 | ctx.mark_non_differentiable(samples) 22 | return samples 23 | 24 | @staticmethod 25 | def backward(ctx, grad_output): 26 | return None 27 | 28 | 29 | class Normal(Function): 30 | @staticmethod 31 | def forward(ctx, means, stddevs=None): 32 | samples = torch.normal(means, stddevs) 33 | ctx.mark_non_differentiable(samples) 34 | return samples 35 | 36 | @staticmethod 37 | def backward(ctx, grad_output): 38 | return None, None 39 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/backends/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/backends/__init__.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/contrib/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/contrib/__init__.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/cuda/error.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/cuda/error.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/cuda/nvtx.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | 3 | lib = None 4 | 5 | __all__ = ['range_push', 'range_pop', 'mark'] 6 | 7 | 8 | def _libnvToolsExt(): 9 | global lib 10 | if lib is None: 11 | lib = ctypes.cdll.LoadLibrary(None) 12 | lib.nvtxMarkA.restype = None 13 | return lib 14 | 15 | 16 | def range_push(msg): 17 | """ 18 | 设置一个固定范围的堆栈,返回的堆栈范围深度从0开始. 19 | 20 | Arguments: 21 | msg (string): 范围(用 ASCII 编码设置) 22 | """ 23 | if _libnvToolsExt() is None: 24 | raise RuntimeError('Unable to load nvToolsExt library') 25 | return lib.nvtxRangePushA(ctypes.c_char_p(msg.encode("ascii"))) 26 | 27 | 28 | def range_pop(): 29 | """ 30 | 弹出一个固定范围的堆栈,返回的堆栈范围深度从0结束. 31 | """ 32 | if _libnvToolsExt() is None: 33 | raise RuntimeError('Unable to load nvToolsExt library') 34 | return lib.nvtxRangePop() 35 | 36 | 37 | def mark(msg): 38 | """ 39 | 描述在某个时刻发生的瞬间事件. 40 | 41 | Arguments: 42 | msg (string): 事件(用 ASCII 编码表示). 43 | """ 44 | if _libnvToolsExt() is None: 45 | raise RuntimeError('Unable to load nvToolsExt library') 46 | return lib.nvtxMarkA(ctypes.c_char_p(msg.encode("ascii"))) 47 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/cuda/profiler.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tempfile 3 | import contextlib 4 | from . import cudart, check_error 5 | 6 | 7 | class cudaOutputMode(object): 8 | cudaKeyValuePair = ctypes.c_int(0) 9 | cudaCSV = ctypes.c_int(1) 10 | 11 | @staticmethod 12 | def for_key(key): 13 | if key == 'key_value': 14 | return cudaOutputMode.cudaKeyValuePair 15 | elif key == 'csv': 16 | return cudaOutputMode.cudaCSV 17 | else: 18 | raise RuntimeError("supported CUDA profiler output modes are: key_value and csv") 19 | 20 | DEFAULT_FLAGS = [ 21 | "gpustarttimestamp", 22 | "gpuendtimestamp", 23 | "gridsize3d", 24 | "threadblocksize", 25 | "streamid", 26 | "enableonstart 0", 27 | "conckerneltrace", 28 | ] 29 | 30 | 31 | def init(output_file, flags=None, output_mode='key_value'): 32 | flags = DEFAULT_FLAGS if flags is None else flags 33 | output_mode = cudaOutputMode.for_key(output_mode) 34 | with tempfile.NamedTemporaryFile(delete=True) as f: 35 | f.write(b'\n'.join(map(lambda f: f.encode('ascii'), flags))) 36 | f.flush() 37 | check_error(cudart().cudaProfilerInitialize( 38 | ctypes.c_char_p(f.name.encode('ascii')), ctypes.c_char_p(output_file.encode('ascii')), output_mode)) 39 | 40 | 41 | def start(): 42 | check_error(cudart().cudaProfilerStart()) 43 | 44 | 45 | def stop(): 46 | check_error(cudart().cudaProfilerStop()) 47 | 48 | 49 | @contextlib.contextmanager 50 | def profile(): 51 | try: 52 | start() 53 | yield 54 | finally: 55 | stop() 56 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/for_onnx/__init__.py: -------------------------------------------------------------------------------- 1 | from .onnx import * 2 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/jit/passes/__init__.py: -------------------------------------------------------------------------------- 1 | from .inplace import _check_inplace 2 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/jit/passes/inplace.py: -------------------------------------------------------------------------------- 1 | 2 | def _check_inplace(trace): 3 | """Checks that all PythonOps that were not translated into JIT format are out of place. 4 | 5 | Should be run after the ONNX pass. 6 | """ 7 | graph = trace.graph() 8 | for node in graph.nodes(): 9 | if node.kind() == 'PythonOp': 10 | if node.i('inplace'): 11 | raise RuntimeError("inplace {} not supported in the JIT".format(node.pyname())) 12 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/__init__.py: -------------------------------------------------------------------------------- 1 | """包含从 Lua torch 移植的代码的包. 2 | 3 | 为了能够与现有的模型一起工作, 并简化当前 Lua torch 用户的过渡, 我们特意创建了这个包. 4 | 您可以在 ``torch.legacy.nn`` 中找到 ``nn`` 代码, 并在 ``torch.legacy.optim`` 中进行 ``optim` 优化. 5 | 该 API 应该完适配 Lua torch. 6 | """ 7 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Abs.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Abs(Module): 6 | 7 | def __init__(self): 8 | super(Abs, self).__init__() 9 | 10 | def updateOutput(self, input): 11 | self._backend.Abs_updateOutput( 12 | self._backend.library_state, 13 | input, 14 | self.output 15 | ) 16 | return self.output 17 | 18 | def updateGradInput(self, input, gradOutput): 19 | self._backend.Abs_updateGradInput( 20 | self._backend.library_state, 21 | input, 22 | gradOutput, 23 | self.gradInput 24 | ) 25 | return self.gradInput 26 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/AbsCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class AbsCriterion(Criterion): 6 | 7 | def __init__(self, sizeAverage=True): 8 | super(AbsCriterion, self).__init__() 9 | self.sizeAverage = sizeAverage 10 | self.output_tensor = torch.Tensor(1) 11 | 12 | def updateOutput(self, input, target): 13 | if self.output_tensor is None: 14 | self.output_tensor = input.new(1) 15 | self._backend.AbsCriterion_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | target, 19 | self.output_tensor, 20 | self.sizeAverage, 21 | True, # reduce 22 | ) 23 | self.output = self.output_tensor[0] 24 | return self.output 25 | 26 | def updateGradInput(self, input, target): 27 | implicit_gradOutput = torch.ones(1).type_as(input) 28 | self._backend.AbsCriterion_updateGradInput( 29 | self._backend.library_state, 30 | input, 31 | target, 32 | implicit_gradOutput, 33 | self.gradInput, 34 | self.sizeAverage, 35 | True, # reduce 36 | ) 37 | return self.gradInput 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/AddConstant.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class AddConstant(Module): 6 | 7 | def __init__(self, constant_scalar, inplace=False): 8 | super(AddConstant, self).__init__() 9 | self.constant_scalar = constant_scalar 10 | self.inplace = inplace 11 | 12 | def updateOutput(self, input): 13 | if self.inplace: 14 | input.add_(self.constant_scalar) 15 | self.output.set_(input) 16 | else: 17 | self.output.resize_as_(input) 18 | self.output.copy_(input) 19 | self.output.add_(self.constant_scalar) 20 | 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | if self.inplace: 25 | self.gradInput.set_(gradOutput) 26 | # restore previous input value 27 | input.add_(-self.constant_scalar) 28 | else: 29 | self.gradInput.resize_as_(gradOutput) 30 | self.gradInput.copy_(gradOutput) 31 | 32 | return self.gradInput 33 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/CAddTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class CAddTable(Module): 6 | 7 | def __init__(self, inplace=False): 8 | super(CAddTable, self).__init__() 9 | self.inplace = inplace 10 | self.gradInput = [] 11 | 12 | def updateOutput(self, input): 13 | if self.inplace: 14 | self.output.set_(input[0]) 15 | else: 16 | self.output.resize_as_(input[0]).copy_(input[0]) 17 | 18 | for i in range(1, len(input)): 19 | self.output.add_(input[i]) 20 | 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | for i in range(len(input)): 25 | if i >= len(self.gradInput): 26 | assert i == len(self.gradInput) 27 | self.gradInput.append(input[0].new()) 28 | 29 | if self.inplace: 30 | self.gradInput[i].set_(gradOutput) 31 | else: 32 | self.gradInput[i].resize_as_(input[i]).copy_(gradOutput) 33 | 34 | del self.gradInput[len(input):] 35 | 36 | return self.gradInput 37 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/CDivTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class CDivTable(Module): 6 | 7 | def __init__(self, ): 8 | super(CDivTable, self).__init__() 9 | self.gradInput = [] 10 | 11 | def updateOutput(self, input): 12 | self.output.resize_as_(input[0]).copy_(input[0]) 13 | self.output.div_(input[1]) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | while len(self.gradInput) < 2: 18 | self.gradInput.append(input[0].new()) 19 | self.gradInput[0].resize_as_(input[0]).copy_(gradOutput, broadcast=False).div_(input[1]) 20 | self.gradInput[1].resize_as_(input[1]).zero_().addcdiv_(-1, self.gradInput[0], input[1]).mul_(input[0]) 21 | 22 | del self.gradInput[len(input):] 23 | 24 | return self.gradInput 25 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/CSubTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class CSubTable(Module): 6 | 7 | def __init__(self, ): 8 | super(CSubTable, self).__init__() 9 | self.gradInput = [torch.Tensor(), torch.Tensor()] 10 | 11 | def updateOutput(self, input): 12 | self.output.resize_as_(input[0]).copy_(input[0]) 13 | self.output.add_(-1, input[1]) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | if self.gradInput[0] is None: 18 | self.gradInput[0] = input[0].new() 19 | if self.gradInput[1] is None: 20 | self.gradInput[1] = input[1].new() 21 | self.gradInput[0].resize_as_(input[0]).copy_(gradOutput) 22 | self.gradInput[1].resize_as_(input[1]).copy_(gradOutput).mul_(-1) 23 | 24 | self.gradInput = self.gradInput[:2] 25 | return self.gradInput 26 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Clamp.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .HardTanh import HardTanh 3 | 4 | 5 | class Clamp(HardTanh): 6 | 7 | def __init__(self, min_value, max_value): 8 | super(Clamp, self,).__init__(min_value, max_value) 9 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Contiguous.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Contiguous(Module): 6 | 7 | def updateOutput(self, input): 8 | if not input.is_contiguous(): 9 | self.output.resize_as_(input).copy_(input) 10 | else: 11 | self.output.set_(input) 12 | 13 | return self.output 14 | 15 | def updateGradInput(self, input, gradOutput): 16 | if not gradOutput.is_contiguous(): 17 | self.gradInput.resize_as_(gradOutput).copy_(gradOutput) 18 | else: 19 | self.gradInput.set_(gradOutput) 20 | 21 | return self.gradInput 22 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Copy.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Copy(Module): 6 | 7 | def __init__(self, intype, outtype, dontCast=False): 8 | self.dontCast = dontCast 9 | super(Copy, self).__init__() 10 | self.gradInput = intype() 11 | self.output = outtype() 12 | 13 | def updateOutput(self, input): 14 | self.output.resize_(input.size()).copy_(input) 15 | return self.output 16 | 17 | def updateGradInput(self, input, gradOutput): 18 | self.gradInput.resize_(gradOutput.size()).copy_(gradOutput) 19 | return self.gradInput 20 | 21 | def type(self, type=None, tensorCache=None): 22 | if type and self.dontCast: 23 | return self 24 | 25 | return super(Copy, self).type(self, type, tensorCache) 26 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Criterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import recursiveType 4 | import torch._thnn 5 | 6 | 7 | class Criterion(object): 8 | 9 | def __init__(self): 10 | self.gradInput = torch.Tensor() 11 | self.output = 0 12 | self._backend = torch._thnn.type2backend[type(self.gradInput)] 13 | 14 | def updateOutput(self, input, target): 15 | raise NotImplementedError 16 | 17 | def forward(self, input, target): 18 | return self.updateOutput(input, target) 19 | 20 | def backward(self, input, target): 21 | return self.updateGradInput(input, target) 22 | 23 | def updateGradInput(self, input, target): 24 | raise NotImplementedError 25 | 26 | def clone(self): 27 | raise NotImplementedError 28 | 29 | def type(self, type, tensorCache=None): 30 | # find all tensors and convert them 31 | for key, param in self.__dict__.items(): 32 | setattr(self, key, recursiveType(param, type, tensorCache or {})) 33 | 34 | self._backend = torch._thnn.type2backend[type] 35 | return self 36 | 37 | def float(self): 38 | return self.type('torch.FloatTensor') 39 | 40 | def double(self): 41 | return self.type('torch.DoubleTensor') 42 | 43 | def cuda(self): 44 | return self.type('torch.cuda.FloatTensor') 45 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/CriterionTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class CriterionTable(Module): 6 | 7 | def __init__(self, criterion): 8 | super(CriterionTable, self).__init__() 9 | self.criterion = criterion 10 | self.gradInput = [criterion.gradInput] 11 | 12 | def updateOutput(self, input): 13 | self.output = self.criterion.updateOutput(*input) 14 | return self.output 15 | 16 | def updateGradInput(self, input, grad_output): 17 | self.criterion.updateGradInput(*input) 18 | return self.gradInput 19 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/CrossEntropyCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | from .LogSoftMax import LogSoftMax 4 | from .ClassNLLCriterion import ClassNLLCriterion 5 | 6 | 7 | class CrossEntropyCriterion(Criterion): 8 | 9 | def __init__(self, weights=None): 10 | super(CrossEntropyCriterion, self).__init__() 11 | self.lsm = LogSoftMax() 12 | self.nll = ClassNLLCriterion(weights) 13 | 14 | def updateOutput(self, input, target): 15 | input = input.squeeze() 16 | target = target.squeeze() 17 | self.lsm.updateOutput(input) 18 | self.nll.updateOutput(self.lsm.output, target) 19 | self.output = self.nll.output 20 | return self.output 21 | 22 | def updateGradInput(self, input, target): 23 | size = input.size() 24 | input = input.squeeze() 25 | target = target.squeeze() 26 | self.nll.updateGradInput(self.lsm.output, target) 27 | self.lsm.updateGradInput(input, self.nll.gradInput) 28 | self.gradInput = self.lsm.gradInput.view(size) 29 | return self.gradInput 30 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/DistKLDivCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class DistKLDivCriterion(Criterion): 6 | 7 | def __init__(self, sizeAverage=True): 8 | super(DistKLDivCriterion, self).__init__() 9 | self.sizeAverage = sizeAverage 10 | self.output_tensor = torch.Tensor(1) 11 | 12 | def updateOutput(self, input, target): 13 | assert input.is_same_size(target) 14 | if self.output_tensor is None: 15 | self.output_tensor = input.new(1) 16 | self._backend.DistKLDivCriterion_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | target, 20 | self.output_tensor, 21 | self.sizeAverage, 22 | True, # reduce 23 | ) 24 | self.output = self.output_tensor[0] 25 | return self.output 26 | 27 | def updateGradInput(self, input, target): 28 | assert input.is_same_size(target) 29 | implicit_gradOutput = torch.ones(1).type_as(input) 30 | self._backend.DistKLDivCriterion_updateGradInput( 31 | self._backend.library_state, 32 | input, 33 | target, 34 | implicit_gradOutput, 35 | self.gradInput, 36 | self.sizeAverage, 37 | True, # reduce 38 | ) 39 | return self.gradInput 40 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/DotProduct.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class DotProduct(Module): 7 | 8 | def __init__(self): 9 | super(DotProduct, self).__init__() 10 | self.gradInput = [torch.Tensor(), torch.Tensor()] 11 | self.buffer = None 12 | 13 | def updateOutput(self, input): 14 | input1, input2 = input[0], input[1] 15 | 16 | if self.buffer is None: 17 | self.buffer = input1.new() 18 | 19 | torch.mul(input1, input2, out=self.buffer) 20 | torch.sum(self.buffer, 1, True, out=self.output) 21 | self.output.resize_(input1.size(0)) 22 | return self.output 23 | 24 | def updateGradInput(self, input, gradOutput): 25 | v1 = input[0] 26 | v2 = input[1] 27 | not_batch = False 28 | 29 | if len(self.gradInput) != 2: 30 | if self.gradInput[0] is None: 31 | self.gradInput[0] = input[0].new() 32 | if self.gradInput[1] is None: 33 | self.gradInput[1] = input[1].new() 34 | self.gradInput = self.gradInput[:2] 35 | 36 | gw1 = self.gradInput[0] 37 | gw2 = self.gradInput[1] 38 | gw1.resize_as_(v1).copy_(v2) 39 | gw2.resize_as_(v2).copy_(v1) 40 | 41 | go = gradOutput.contiguous().view(-1, 1).expand_as(v1) 42 | gw1.mul_(go) 43 | gw2.mul_(go) 44 | 45 | return self.gradInput 46 | 47 | def clearState(self): 48 | clear(self, 'buffer') 49 | return super(DotProduct, self).clearState() 50 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Dropout.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class Dropout(Module): 7 | 8 | def __init__(self, p=0.5, inplace=False): 9 | super(Dropout, self).__init__() 10 | self.p = p 11 | self.inplace = inplace 12 | self.train = True 13 | self.noise = torch.Tensor() 14 | 15 | def updateOutput(self, input): 16 | if self.inplace: 17 | self.output.set_(input) 18 | else: 19 | self.output.resize_as_(input).copy_(input) 20 | 21 | if self.p > 0 and self.train: 22 | self.noise.resize_as_(input) 23 | self.noise.bernoulli_(1 - self.p) 24 | self.noise.div_(1 - self.p) 25 | self.output.mul_(self.noise) 26 | 27 | return self.output 28 | 29 | def updateGradInput(self, input, gradOutput): 30 | if self.inplace: 31 | self.gradInput.set_(gradOutput) 32 | else: 33 | self.gradInput.resize_as_(gradOutput).copy_(gradOutput) 34 | 35 | if self.p > 0 and self.train: 36 | self.gradInput.mul_(self.noise) # simply mask the gradients with the noise vector 37 | 38 | return self.gradInput 39 | 40 | def setp(self, p): 41 | self.p = p 42 | 43 | def __repr__(self): 44 | return super(Dropout, self).__repr__() + '({:.4f})'.format(self.p) 45 | 46 | def clearState(self): 47 | clear(self, 'noise') 48 | return super(Dropout, self).clearState() 49 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/ELU.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf8 -*- 2 | import torch 3 | from .Module import Module 4 | 5 | 6 | class ELU(Module): 7 | """ 8 | Djork-Arné Clevert, Thomas Unterthiner, Sepp Hochreiter 9 | Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) 10 | http.//arxiv.org/pdf/1511.07289.pdf 11 | """ 12 | 13 | def __init__(self, alpha=1., inplace=False): 14 | assert type(alpha) == float 15 | super(ELU, self).__init__() 16 | self.alpha = alpha 17 | self.inplace = inplace 18 | 19 | def updateOutput(self, input): 20 | self._backend.ELU_updateOutput( 21 | self._backend.library_state, 22 | input, 23 | self.output, 24 | self.alpha, 25 | self.inplace 26 | ) 27 | return self.output 28 | 29 | def updateGradInput(self, input, gradOutput): 30 | self._backend.ELU_updateGradInput( 31 | self._backend.library_state, 32 | gradOutput, 33 | self.gradInput, 34 | self.output, 35 | self.alpha, 36 | self.inplace 37 | ) 38 | return self.gradInput 39 | 40 | def __repr__(self): 41 | return '{}(alpha={:.3f})'.format(str(type(self)), self.alpha) 42 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Exp.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Exp(Module): 6 | 7 | def updateOutput(self, input): 8 | return torch.exp(input, out=self.output) 9 | 10 | def updateGradInput(self, input, gradOutput): 11 | return torch.mul(self.output, gradOutput, out=self.gradInput) 12 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/GradientReversal.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class GradientReversal(Module): 6 | 7 | def __init__(self, lambd=1): 8 | super(GradientReversal, self).__init__() 9 | self.lambd = lambd 10 | 11 | def setLambda(self, lambd): 12 | self.lambd = lambd 13 | 14 | def updateOutput(self, input): 15 | self.output.set_(input) 16 | return self.output 17 | 18 | def updateGradInput(self, input, gradOutput): 19 | self.gradInput.resize_as_(gradOutput) 20 | self.gradInput.copy_(gradOutput) 21 | self.gradInput.mul_(-self.lambd) 22 | return self.gradInput 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/HardShrink.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class HardShrink(Module): 6 | 7 | def __init__(self, lambd=0.5): 8 | assert type(lambd) == float 9 | super(HardShrink, self).__init__() 10 | self.lambd = lambd 11 | 12 | def updateOutput(self, input): 13 | self._backend.HardShrink_updateOutput( 14 | self._backend.library_state, 15 | input, 16 | self.output, 17 | self.lambd 18 | ) 19 | return self.output 20 | 21 | def updateGradInput(self, input, gradOutput): 22 | self._backend.HardShrink_updateGradInput( 23 | self._backend.library_state, 24 | input, 25 | gradOutput, 26 | self.gradInput, 27 | self.lambd 28 | ) 29 | return self.gradInput 30 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/HardTanh.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class HardTanh(Module): 6 | 7 | def __init__(self, min_value=-1, max_value=1, inplace=False): 8 | super(HardTanh, self).__init__() 9 | self.min_val = min_value 10 | self.max_val = max_value 11 | self.inplace = inplace 12 | assert self.max_val > self.min_val 13 | 14 | def updateOutput(self, input): 15 | self._backend.HardTanh_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | self.output, 19 | self.min_val, 20 | self.max_val, 21 | self.inplace 22 | ) 23 | return self.output 24 | 25 | def updateGradInput(self, input, gradOutput): 26 | self._backend.HardTanh_updateGradInput( 27 | self._backend.library_state, 28 | input, 29 | gradOutput, 30 | self.gradInput, 31 | self.min_val, 32 | self.max_val, 33 | self.inplace 34 | ) 35 | return self.gradInput 36 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/HingeEmbeddingCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class HingeEmbeddingCriterion(Criterion): 6 | 7 | def __init__(self, margin=1, sizeAverage=True): 8 | super(HingeEmbeddingCriterion, self).__init__() 9 | self.margin = margin 10 | self.sizeAverage = sizeAverage 11 | self.buffer = None 12 | 13 | def updateOutput(self, input, y): 14 | if self.buffer is None: 15 | self.buffer = input.new() 16 | self.buffer.resize_as_(input).copy_(input) 17 | self.buffer[torch.eq(y, -1.)] = 0 18 | self.output = self.buffer.sum() 19 | 20 | self.buffer.fill_(self.margin).add_(-1, input) 21 | self.buffer.clamp_(min=0) 22 | self.buffer[torch.eq(y, 1.)] = 0 23 | self.output = self.output + self.buffer.sum() 24 | 25 | if self.sizeAverage: 26 | self.output = self.output / input.nelement() 27 | 28 | return self.output 29 | 30 | def updateGradInput(self, input, y): 31 | self.gradInput.resize_as_(input).copy_(y) 32 | self.gradInput[torch.mul(torch.eq(y, -1), torch.gt(input, self.margin))] = 0 33 | 34 | if self.sizeAverage: 35 | self.gradInput.mul_(1. / input.nelement()) 36 | 37 | return self.gradInput 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Identity.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class Identity(Module): 7 | 8 | def updateOutput(self, input): 9 | self.output = input 10 | return self.output 11 | 12 | def updateGradInput(self, input, gradOutput): 13 | self.gradInput = gradOutput 14 | return self.gradInput 15 | 16 | def clearState(self): 17 | clear(self, 'gradInput') 18 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Index.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Index(Module): 6 | 7 | def __init__(self, dimension): 8 | super(Index, self).__init__() 9 | self.dimension = dimension 10 | self.gradInput = [self.gradInput] 11 | 12 | def updateOutput(self, input): 13 | t = input[0] 14 | index = input[1] 15 | torch.index_select(t, self.dimension, index, out=self.output) 16 | return self.output 17 | 18 | def updateGradInput(self, input, gradOutput): 19 | t = input[0] 20 | index = input[1] 21 | 22 | gradInput = self.gradInput[0] # no gradient for the index variable 23 | gradInput.resize_as_(t).zero_() 24 | gradInput.index_add_(self.dimension, index, gradOutput) 25 | return self.gradInput 26 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/L1Cost.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | from .utils import clear 4 | 5 | 6 | class L1Cost(Criterion): 7 | 8 | def __init__(self): 9 | super(L1Cost, self).__init__() 10 | self.output_tensor = torch.Tensor(1) 11 | 12 | def updateOutput(self, input, target=None): 13 | assert target is None 14 | if self.output_tensor is None: 15 | self.output_tensor = input.new(1) 16 | self._backend.L1Cost_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | self.output_tensor 20 | ) 21 | self.output = self.output_tensor[0] 22 | return self.output 23 | 24 | def updateGradInput(self, input, target=None): 25 | assert target is None 26 | self._backend.L1Cost_updateGradInput( 27 | self._backend.library_state, 28 | input, 29 | None, 30 | self.gradInput 31 | ) 32 | return self.gradInput 33 | 34 | def clearState(self): 35 | clear(self, 'output_tensor') 36 | return super(L1Cost, self).clearState() 37 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/L1HingeEmbeddingCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class L1HingeEmbeddingCriterion(Criterion): 6 | 7 | def __init__(self, margin=1): 8 | super(L1HingeEmbeddingCriterion, self).__init__() 9 | self.margin = margin 10 | self.gradInput = [torch.Tensor(), torch.Tensor()] 11 | 12 | def updateOutput(self, input, y): 13 | self.output = input[0].dist(input[1], 1) 14 | if y == -1: 15 | self.output = max(0, self.margin - self.output) 16 | 17 | return self.output 18 | 19 | def _mathsign(t): 20 | return 1 if x > 0 else -1 21 | 22 | def updateGradInput(self, input, y): 23 | self.gradInput[0].resize_as_(input[0]) 24 | self.gradInput[1].resize_as_(input[1]) 25 | self.gradInput[0].copy_(input[0]) 26 | self.gradInput[0].add_(-1, input[1]) 27 | dist = self.gradInput[0].norm(1) 28 | self.gradInput[0].sign_() 29 | if y == -1: # just to avoid a mul by 1 30 | if dist > self.margin: 31 | self.gradInput[0].zero_() 32 | else: 33 | self.gradInput[0].mul_(-1) 34 | 35 | self.gradInput[1].zero_().add_(-1, self.gradInput[0]) 36 | return self.gradInput 37 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/L1Penalty.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | # This module acts as an L1 latent state regularizer, adding the 5 | # [gradOutput] to the gradient of the L1 loss. The [input] is copied to 6 | # the [output]. 7 | 8 | 9 | class L1Penalty(Module): 10 | 11 | def __init__(self, l1weight, sizeAverage=False, provideOutput=True): 12 | super(L1Penalty, self).__init__() 13 | self.l1weight = l1weight 14 | self.sizeAverage = sizeAverage 15 | self.provideOutput = provideOutput 16 | 17 | def updateOutput(self, input): 18 | m = self.l1weight 19 | if self.sizeAverage: 20 | m = m / input.nelement() 21 | 22 | loss = m * input.norm(1) 23 | self.loss = loss 24 | self.output = input 25 | return self.output 26 | 27 | def updateGradInput(self, input, gradOutput): 28 | m = self.l1weight 29 | if self.sizeAverage: 30 | m = m / input.nelement() 31 | 32 | self.gradInput.resize_as_(input).copy_(input).sign_().mul_(m) 33 | 34 | if self.provideOutput: 35 | self.gradInput.add_(gradOutput) 36 | 37 | return self.gradInput 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/LeakyReLU.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class LeakyReLU(Module): 6 | 7 | def __init__(self, negval=1 / 100, inplace=False): 8 | super(LeakyReLU, self).__init__() 9 | if isinstance(negval, bool): 10 | inplace = negval 11 | self.negval = 1 / 100 12 | else: 13 | self.negval = negval 14 | 15 | # default for inplace is False 16 | self.inplace = inplace 17 | if self.negval < 0: 18 | # TODO: warning here 19 | self.inplace = False 20 | 21 | def updateOutput(self, input): 22 | self._backend.LeakyReLU_updateOutput( 23 | self._backend.library_state, 24 | input, 25 | self.output, 26 | self.negval, 27 | self.inplace 28 | ) 29 | return self.output 30 | 31 | def updateGradInput(self, input, gradOutput): 32 | self._backend.LeakyReLU_updateGradInput( 33 | self._backend.library_state, 34 | input, 35 | gradOutput, 36 | self.gradInput, 37 | self.negval, 38 | self.inplace 39 | ) 40 | return self.gradInput 41 | 42 | def __repr__(self): 43 | return str(type(self)) + '({:.4f})'.format(self.negval) 44 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Log.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Log(Module): 6 | 7 | def updateOutput(self, input): 8 | self.output.resize_as_(input) 9 | self.output.copy_(input) 10 | self.output.log_() 11 | return self.output 12 | 13 | def updateGradInput(self, input, gradOutput): 14 | self.gradInput.resize_as_(input) 15 | self.gradInput.fill_(1) 16 | self.gradInput.div_(input) 17 | self.gradInput.mul_(gradOutput) 18 | return self.gradInput 19 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/LogSigmoid.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class LogSigmoid(Module): 7 | 8 | def __init__(self): 9 | super(LogSigmoid, self).__init__() 10 | self.buffer = None 11 | 12 | def updateOutput(self, input): 13 | if self.buffer is None: 14 | self.buffer = input.new() 15 | self._backend.LogSigmoid_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | self.output, 19 | self.buffer 20 | ) 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | self._backend.LogSigmoid_updateGradInput( 25 | self._backend.library_state, 26 | input, 27 | gradOutput, 28 | self.gradInput, 29 | self.buffer 30 | ) 31 | return self.gradInput 32 | 33 | def clearState(self): 34 | clear(self, 'buffer') 35 | return super(LogSigmoid, self).clearState() 36 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/LogSoftMax.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class LogSoftMax(Module): 6 | 7 | def __init__(self, dim=None): 8 | super(LogSoftMax, self).__init__() 9 | if dim is not None: 10 | self.dim = dim 11 | 12 | def _get_dim(self, input): 13 | return getattr(self, 'dim', 0 if input.dim() == 1 or input.dim() == 3 else 1) 14 | 15 | def updateOutput(self, input): 16 | self._backend.LogSoftMax_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | self.output, 20 | self._get_dim(input) 21 | ) 22 | return self.output 23 | 24 | def updateGradInput(self, input, gradOutput): 25 | self._backend.LogSoftMax_updateGradInput( 26 | self._backend.library_state, 27 | input, 28 | gradOutput, 29 | self.gradInput, 30 | self.output, 31 | self._get_dim(input) 32 | ) 33 | return self.gradInput 34 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MSECriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class MSECriterion(Criterion): 6 | 7 | def __init__(self, sizeAverage=True): 8 | super(MSECriterion, self).__init__() 9 | self.sizeAverage = sizeAverage 10 | self.output_tensor = None 11 | 12 | def updateOutput(self, input, target): 13 | if self.output_tensor is None: 14 | self.output_tensor = input.new(1) 15 | self._backend.MSECriterion_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | target, 19 | self.output_tensor, 20 | self.sizeAverage, 21 | True, # reduce 22 | ) 23 | self.output = self.output_tensor[0] 24 | return self.output 25 | 26 | def updateGradInput(self, input, target): 27 | implicit_gradOutput = torch.Tensor([1]).type(input.type()) 28 | 29 | self._backend.MSECriterion_updateGradInput( 30 | self._backend.library_state, 31 | input, 32 | target, 33 | implicit_gradOutput, 34 | self.gradInput, 35 | self.sizeAverage, 36 | True, # reduce 37 | ) 38 | return self.gradInput 39 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MarginCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class MarginCriterion(Criterion): 6 | 7 | def __init__(self, margin=1, sizeAverage=True): 8 | super(MarginCriterion, self).__init__() 9 | self.sizeAverage = True 10 | self.margin = margin 11 | self.output_tensor = None 12 | 13 | def updateOutput(self, input, target): 14 | if self.output_tensor is None: 15 | self.output_tensor = input.new(1) 16 | self._backend.MarginCriterion_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | target, 20 | self.output_tensor, 21 | self.sizeAverage, 22 | self.margin 23 | ) 24 | self.output = self.output_tensor[0] 25 | return self.output 26 | 27 | def updateGradInput(self, input, target): 28 | self._backend.MarginCriterion_updateGradInput( 29 | self._backend.library_state, 30 | input, 31 | target, 32 | self.gradInput, 33 | self.sizeAverage, 34 | self.margin 35 | ) 36 | return self.gradInput 37 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Mean.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Sum import Sum 3 | 4 | """ 5 | 6 | This file is still here because of backward compatibility. 7 | 8 | Please use instead "nn.Sum(dimension, nInputDims, sizeAverage)" 9 | 10 | """ 11 | 12 | 13 | class Mean(Sum): 14 | 15 | def __init__(self, dimension): 16 | super(Mean, self).__init__(dimension, True) 17 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Mul.py: -------------------------------------------------------------------------------- 1 | import math 2 | import torch 3 | from .Module import Module 4 | 5 | 6 | class Mul(Module): 7 | 8 | def __init__(self): 9 | super(Mul, self).__init__() 10 | self.weight = torch.Tensor(1) 11 | self.gradWeight = torch.Tensor(1) 12 | self.reset() 13 | 14 | def reset(self, stdv=None): 15 | if stdv is not None: 16 | stdv = stdv * math.sqrt(3) 17 | else: 18 | stdv = 1. / math.sqrt(self.weight.size(0)) 19 | self.weight.uniform_(-stdv, stdv) 20 | 21 | def updateOutput(self, input): 22 | self.output.resize_as_(input).copy_(input) 23 | self.output.mul_(self.weight[0]) 24 | return self.output 25 | 26 | def updateGradInput(self, input, gradOutput): 27 | self.gradInput.resize_as_(input).zero_() 28 | self.gradInput.add_(self.weight[0], gradOutput) 29 | return self.gradInput 30 | 31 | def accGradParameters(self, input, gradOutput, scale=1): 32 | self.gradWeight[0] = (self.gradWeight[0] + 33 | scale * input.contiguous().view(-1).dot(gradOutput.contiguous().view(-1))) 34 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MulConstant.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class MulConstant(Module): 6 | 7 | def __init__(self, constant_scalar, inplace=False): 8 | super(MulConstant, self).__init__() 9 | self.constant_scalar = constant_scalar 10 | self.inplace = inplace 11 | 12 | def updateOutput(self, input): 13 | if self.inplace: 14 | input.mul_(self.constant_scalar) 15 | self.output.set_(input) 16 | else: 17 | self.output.resize_as_(input) 18 | self.output.copy_(input) 19 | self.output.mul_(self.constant_scalar) 20 | 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | if self.gradInput is None: 25 | return 26 | 27 | if self.inplace: 28 | gradOutput.mul_(self.constant_scalar) 29 | self.gradInput.set_(gradOutput) 30 | # restore previous input value 31 | input.div_(self.constant_scalar) 32 | else: 33 | self.gradInput.resize_as_(gradOutput) 34 | self.gradInput.copy_(gradOutput) 35 | self.gradInput.mul_(self.constant_scalar) 36 | 37 | return self.gradInput 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MultiCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | from .utils import recursiveResizeAs, recursiveFill, recursiveAdd 4 | 5 | 6 | class MultiCriterion(Criterion): 7 | 8 | def __init__(self, ): 9 | super(MultiCriterion, self).__init__() 10 | self.criterions = [] 11 | self.weights = torch.DoubleStorage() 12 | 13 | def add(self, criterion, weight=1): 14 | self.criterions.append(criterion) 15 | new_weights = torch.DoubleStorage(len(self.criterions)) 16 | for i, v in enumerate(self.weights): 17 | new_weights[i] = v 18 | new_weights[len(self.criterions) - 1] = weight 19 | self.weights = new_weights 20 | return self 21 | 22 | def updateOutput(self, input, target): 23 | self.output = 0 24 | for i in range(len(self.criterions)): 25 | self.output = self.output + self.weights[i] * self.criterions[i].updateOutput(input, target) 26 | 27 | return self.output 28 | 29 | def updateGradInput(self, input, target): 30 | self.gradInput = recursiveResizeAs(self.gradInput, input)[0] 31 | recursiveFill(self.gradInput, 0) 32 | for i in range(len(self.criterions)): 33 | recursiveAdd(self.gradInput, self.weights[i], self.criterions[i].updateGradInput(input, target)) 34 | 35 | return self.gradInput 36 | 37 | def type(self, type): 38 | for criterion in self.criterions: 39 | criterion.type(type) 40 | 41 | return super(MultiCriterion, self).type(type) 42 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MultiLabelMarginCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class MultiLabelMarginCriterion(Criterion): 6 | 7 | def __init__(self, sizeAverage=True): 8 | super(MultiLabelMarginCriterion, self).__init__() 9 | self.sizeAverage = sizeAverage 10 | self.isTarget = torch.Tensor() 11 | self.output_tensor = None 12 | 13 | def updateOutput(self, input, target): 14 | if self.output_tensor is None: 15 | self.output_tensor = input.new(1) 16 | target = target.long() 17 | self._backend.MultiLabelMarginCriterion_updateOutput( 18 | self._backend.library_state, 19 | input, 20 | target, 21 | self.output_tensor, 22 | self.isTarget, 23 | self.sizeAverage 24 | ) 25 | self.output = self.output_tensor[0] 26 | return self.output 27 | 28 | def updateGradInput(self, input, target): 29 | target = target.long() 30 | self._backend.MultiLabelMarginCriterion_updateGradInput( 31 | self._backend.library_state, 32 | input, 33 | target, 34 | self.gradInput, 35 | self.isTarget, 36 | self.sizeAverage 37 | ) 38 | return self.gradInput 39 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MultiLabelSoftMarginCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | from .Sigmoid import Sigmoid 4 | from .BCECriterion import BCECriterion 5 | 6 | 7 | class MultiLabelSoftMarginCriterion(Criterion): 8 | """ 9 | A MultiLabel multiclass criterion based on sigmoid: 10 | 11 | the loss is: 12 | l(x, y) = - sum_i y[i] * log(p[i]) + (1 - y[i]) * log (1 - p[i]) 13 | where p[i] = exp(x[i]) / (1 + exp(x[i])) 14 | 15 | and with weights: 16 | l(x, y) = - sum_i weights[i] (y[i] * log(p[i]) + (1 - y[i]) * log (1 - p[i])) 17 | 18 | 19 | """ 20 | 21 | def __init__(self, weights=None): 22 | super(MultiLabelSoftMarginCriterion, self).__init__() 23 | self.lsm = Sigmoid() 24 | self.nll = BCECriterion(weights) 25 | 26 | def updateOutput(self, input, target): 27 | input = input if input.nelement() == 1 else input.squeeze() 28 | target = target if target.nelement() == 1 else target.squeeze() 29 | self.lsm.updateOutput(input) 30 | self.nll.updateOutput(self.lsm.output, target) 31 | self.output = self.nll.output 32 | return self.output 33 | 34 | def updateGradInput(self, input, target): 35 | size = input.size() 36 | input = input if input.nelement() == 1 else input.squeeze() 37 | target = target if target.nelement() == 1 else target.squeeze() 38 | self.nll.updateGradInput(self.lsm.output, target) 39 | self.lsm.updateGradInput(input, self.nll.gradInput) 40 | self.gradInput = self.lsm.gradInput.view(size) 41 | return self.gradInput 42 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/MultiMarginCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class MultiMarginCriterion(Criterion): 6 | 7 | def __init__(self, p=1, weights=None, margin=1, sizeAverage=True): 8 | super(MultiMarginCriterion, self).__init__() 9 | if p != 1 and p != 2: 10 | raise ValueError("only p == 1 and p == 2 supported") 11 | self.p = p 12 | self.margin = margin 13 | self.sizeAverage = sizeAverage 14 | if weights is not None: 15 | assert weights.dim() == 1 16 | self.weights = weights 17 | self.output_tensor = None 18 | 19 | def updateOutput(self, input, target): 20 | if self.output_tensor is None: 21 | self.output_tensor = input.new(1) 22 | target = target.long() 23 | self._backend.MultiMarginCriterion_updateOutput( 24 | self._backend.library_state, 25 | input, 26 | target, 27 | self.output_tensor, 28 | self.sizeAverage, 29 | self.p, 30 | self.weights, 31 | self.margin 32 | ) 33 | self.output = self.output_tensor[0] 34 | return self.output 35 | 36 | def updateGradInput(self, input, target): 37 | target = target.long() 38 | self._backend.MultiMarginCriterion_updateGradInput( 39 | self._backend.library_state, 40 | input, 41 | target, 42 | self.gradInput, 43 | self.sizeAverage, 44 | self.p, 45 | self.weights, 46 | self.margin 47 | ) 48 | return self.gradInput 49 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Narrow.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Narrow(Module): 6 | 7 | def __init__(self, dimension, offset, length=1): 8 | super(Narrow, self).__init__() 9 | self.dimension = dimension 10 | self.index = offset 11 | self.length = length 12 | 13 | def updateOutput(self, input): 14 | length = self.length 15 | if length < 0: 16 | length = input.size(self.dimension) - self.index + self.length + 1 17 | 18 | output = input.narrow(self.dimension, self.index, length) 19 | self.output = self.output.type_as(output) 20 | self.output.resize_as_(output).copy_(output) 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | length = self.length 25 | if length < 0: 26 | length = input.size(self.dimension) - self.index + self.length + 1 27 | 28 | self.gradInput = self.gradInput.type_as(input) 29 | self.gradInput.resize_as_(input).zero_() 30 | self.gradInput.narrow(self.dimension, self.index, length).copy_(gradOutput) 31 | return self.gradInput 32 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/NarrowTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear, recursiveResizeAs, recursiveFill 4 | 5 | 6 | class NarrowTable(Module): 7 | 8 | def __init__(self, offset, length=1): 9 | super(NarrowTable, self).__init__() 10 | self.offset = offset 11 | self.length = length 12 | self.output = [] 13 | self.gradInput = [] 14 | 15 | def updateOutput(self, input): 16 | self.output[:] = [input[self.offset + i] for i in range(self.length)] 17 | return self.output 18 | 19 | def updateGradInput(self, input, gradOutput): 20 | if len(self.gradInput) != len(input): 21 | self.gradInput[:] = [None for i in range(len(input))] 22 | 23 | assert len(gradOutput) == self.length 24 | for i in range(self.length): 25 | self.gradInput[self.offset + i] = gradOutput[i] 26 | 27 | for i in range(len(input)): 28 | if i < self.offset or i >= self.offset + self.length: 29 | gi = self.gradInput[i] 30 | if gi is None: 31 | gi = input[i].new() 32 | self.gradInput[i] = recursiveResizeAs(gi, input[i])[0] 33 | recursiveFill(self.gradInput[i], 0) 34 | 35 | return self.gradInput 36 | 37 | def type(self, type=None, tensorCache=None): 38 | if not type: 39 | return self._type 40 | clear(self, 'output', 'gradInput') 41 | return super(NarrowTable, self).type(self, type, tensorCache) 42 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/PReLU.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class PReLU(Module): 7 | 8 | def __init__(self, nOutputPlane=0): 9 | super(PReLU, self).__init__() 10 | # if no argument provided, use shared model (weight is scalar) 11 | self.nOutputPlane = nOutputPlane 12 | self.weight = torch.Tensor(nOutputPlane or 1).fill_(0.25) 13 | self.gradWeight = torch.Tensor(nOutputPlane or 1) 14 | 15 | def updateOutput(self, input): 16 | self._backend.PReLU_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | self.output, 20 | self.weight 21 | ) 22 | return self.output 23 | 24 | def updateGradInput(self, input, gradOutput): 25 | self._backend.PReLU_updateGradInput( 26 | self._backend.library_state, 27 | input, 28 | gradOutput, 29 | self.gradInput, 30 | self.weight 31 | ) 32 | return self.gradInput 33 | 34 | def accGradParameters(self, input, gradOutput, scale=1): 35 | self._backend.PReLU_accGradParameters( 36 | self._backend.library_state, 37 | input, 38 | gradOutput, 39 | self.gradInput, 40 | self.weight, 41 | self.gradWeight, 42 | scale 43 | ) 44 | return self.gradWeight 45 | 46 | def clearState(self): 47 | clear(self, 'gradWeightBuf', 'gradWeightBuf2') 48 | return super(PReLU, self).clearState() 49 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/ParallelCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | from .utils import recursiveResizeAs, recursiveFill, recursiveAdd 4 | 5 | 6 | class ParallelCriterion(Criterion): 7 | 8 | def __init__(self, repeatTarget=False): 9 | super(ParallelCriterion, self).__init__() 10 | self.criterions = [] 11 | self.weights = [] 12 | self.gradInput = [] 13 | self.repeatTarget = repeatTarget 14 | 15 | def add(self, criterion, weight=1): 16 | self.criterions.append(criterion) 17 | self.weights.append(weight) 18 | return self 19 | 20 | def updateOutput(self, input, target): 21 | self.output = 0 22 | for i, criterion in enumerate(self.criterions): 23 | current_target = target if self.repeatTarget else target[i] 24 | self.output += self.weights[i] * criterion.updateOutput(input[i], current_target) 25 | 26 | return self.output 27 | 28 | def updateGradInput(self, input, target): 29 | self.gradInput = recursiveResizeAs(self.gradInput, input)[0] 30 | recursiveFill(self.gradInput, 0) 31 | for i, criterion in enumerate(self.criterions): 32 | current_target = target if self.repeatTarget else target[i] 33 | recursiveAdd(self.gradInput[i], self.weights[i], criterion.updateGradInput(input[i], current_target)) 34 | 35 | return self.gradInput 36 | 37 | def type(self, type=None, tensorCache=None): 38 | self.gradInput = [] 39 | return super(ParallelCriterion, self).type(type, tensorCache) 40 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Power.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Power(Module): 6 | 7 | def __init__(self, p): 8 | super(Power, self).__init__() 9 | self.pow = p 10 | 11 | def updateOutput(self, input): 12 | self.output.resize_as_(input).copy_(input) 13 | self.output.pow_(self.pow) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | self.gradInput.resize_as_(input).copy_(input) 18 | self.gradInput.pow_(self.pow - 1) 19 | self.gradInput.mul_(gradOutput).mul_(self.pow) 20 | return self.gradInput 21 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/RReLU.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class RReLU(Module): 7 | 8 | def __init__(self, lower=1. / 8, upper=1. / 3, inplace=False): 9 | super(RReLU, self).__init__() 10 | self.lower = lower 11 | self.upper = upper 12 | self.inplace = inplace 13 | 14 | assert self.lower <= self.upper and self.lower >= 0 and self.upper >= 0 15 | self.noise = torch.Tensor() 16 | self.train = True 17 | 18 | def updateOutput(self, input): 19 | self._backend.RReLU_updateOutput( 20 | self._backend.library_state, 21 | input, 22 | self.output, 23 | self.noise, 24 | self.lower, 25 | self.upper, 26 | self.train, 27 | self.inplace, 28 | torch.default_generator if not input.is_cuda else 0 29 | ) 30 | return self.output 31 | 32 | def updateGradInput(self, input, gradOutput): 33 | self._backend.RReLU_updateGradInput( 34 | self._backend.library_state, 35 | input, 36 | gradOutput, 37 | self.gradInput, 38 | self.noise, 39 | self.lower, 40 | self.upper, 41 | self.train, 42 | self.inplace 43 | ) 44 | return self.gradInput 45 | 46 | def __repr__(self): 47 | return super(RReLU, self).__repr__() + '({:.4f}, {:.4f})'.format(self.lower, self.upper) 48 | 49 | def clearState(self): 50 | clear(self, 'noise') 51 | return super(RReLU, self).clearState() 52 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/ReLU.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Threshold import Threshold 3 | 4 | 5 | class ReLU(Threshold): 6 | 7 | def __init__(self, inplace=False): 8 | super(ReLU, self).__init__(0, 0, inplace) 9 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/ReLU6.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class ReLU6(Module): 6 | 7 | def __init__(self, inplace=False): 8 | super(ReLU6, self).__init__() 9 | self.inplace = inplace 10 | 11 | def updateOutput(self, input): 12 | self._backend.HardTanh_updateOutput( 13 | self._backend.library_state, 14 | input, 15 | self.output, 16 | 0, 6, self.inplace 17 | ) 18 | return self.output 19 | 20 | def updateGradInput(self, input, gradOutput): 21 | self._backend.HardTanh_updateGradInput( 22 | self._backend.library_state, 23 | input, 24 | gradOutput, 25 | self.gradInput, 26 | 0, 6, self.inplace 27 | ) 28 | return self.gradInput 29 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Replicate.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Replicate(Module): 6 | 7 | def __init__(self, nf, dim=0): 8 | super(Replicate, self).__init__() 9 | self.nfeatures = nf 10 | self.dim = dim 11 | assert self.dim >= 0 12 | 13 | def updateOutput(self, input): 14 | assert self.dim < input.dim() 15 | 16 | size = list(input.size()) 17 | size.insert(self.dim, self.nfeatures) 18 | 19 | stride = list(input.stride()) 20 | stride.insert(self.dim, 0) 21 | 22 | self.output.set_(input.storage(), input.storage_offset(), 23 | torch.Size(size), tuple(stride)) 24 | return self.output 25 | 26 | def updateGradInput(self, input, gradOutput): 27 | self.gradInput.resize_as_(input).zero_() 28 | size = list(input.size()) 29 | size.insert(self.dim, 1) 30 | 31 | gradInput = self.gradInput.view(*size) 32 | torch.sum(gradOutput, self.dim, True, out=gradInput) 33 | return self.gradInput 34 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Select.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Select(Module): 6 | 7 | def __init__(self, dimension, index): 8 | super(Select, self).__init__() 9 | self.dimension = dimension 10 | self.index = index 11 | 12 | def updateOutput(self, input): 13 | index = self.index if self.index >= 0 else input.size(self.dimension) + self.index 14 | output = input.select(self.dimension, index) 15 | self.output.resize_as_(output) 16 | return self.output.copy_(output) 17 | 18 | def updateGradInput(self, input, gradOutput): 19 | index = self.index if self.index >= 0 else input.size(self.dimension) + self.index 20 | self.gradInput.resize_as_(input) 21 | self.gradInput.zero_() 22 | self.gradInput.select(self.dimension, index).copy_(gradOutput) 23 | return self.gradInput 24 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Sigmoid.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Sigmoid(Module): 6 | 7 | def updateOutput(self, input): 8 | self._backend.Sigmoid_updateOutput( 9 | self._backend.library_state, 10 | input, 11 | self.output 12 | ) 13 | return self.output 14 | 15 | def updateGradInput(self, input, gradOutput): 16 | self._backend.Sigmoid_updateGradInput( 17 | self._backend.library_state, 18 | gradOutput, 19 | self.gradInput, 20 | self.output 21 | ) 22 | return self.gradInput 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SmoothL1Criterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class SmoothL1Criterion(Criterion): 6 | 7 | def __init__(self, sizeAverage=True): 8 | super(SmoothL1Criterion, self).__init__() 9 | self.sizeAverage = sizeAverage 10 | self.output_tensor = None 11 | 12 | def updateOutput(self, input, target): 13 | if self.output_tensor is None: 14 | self.output_tensor = input.new(1) 15 | self._backend.SmoothL1Criterion_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | target, 19 | self.output_tensor, 20 | self.sizeAverage, 21 | True, # reduce 22 | ) 23 | self.output = self.output_tensor[0] 24 | return self.output 25 | 26 | def updateGradInput(self, input, target): 27 | implicit_gradOutput = torch.ones(1).type_as(input) 28 | self._backend.SmoothL1Criterion_updateGradInput( 29 | self._backend.library_state, 30 | input, 31 | target, 32 | implicit_gradOutput, 33 | self.gradInput, 34 | self.sizeAverage, 35 | True, # reduce 36 | ) 37 | return self.gradInput 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftMarginCriterion.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Criterion import Criterion 3 | 4 | 5 | class SoftMarginCriterion(Criterion): 6 | 7 | def __init__(self, ): 8 | super(SoftMarginCriterion, self).__init__() 9 | self.sizeAverage = True 10 | self.output_tensor = None 11 | 12 | def updateOutput(self, input, target): 13 | if self.output_tensor is None: 14 | self.output_tensor = input.new(1) 15 | self._backend.SoftMarginCriterion_updateOutput( 16 | self._backend.library_state, 17 | input, 18 | target, 19 | self.output_tensor, 20 | self.sizeAverage 21 | ) 22 | self.output = self.output_tensor[0] 23 | return self.output 24 | 25 | def updateGradInput(self, input, target): 26 | self._backend.SoftMarginCriterion_updateGradInput( 27 | self._backend.library_state, 28 | input, 29 | target, 30 | self.gradInput, 31 | self.sizeAverage 32 | ) 33 | return self.gradInput 34 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftMax.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class SoftMax(Module): 6 | 7 | def __init__(self, dim=None): 8 | super(SoftMax, self).__init__() 9 | if dim is not None: 10 | self.dim = dim 11 | 12 | def _get_dim(self, input): 13 | return getattr(self, 'dim', 0 if input.dim() == 1 or input.dim() == 3 else 1) 14 | 15 | def updateOutput(self, input): 16 | self._backend.SoftMax_updateOutput( 17 | self._backend.library_state, 18 | input, 19 | self.output, 20 | self._get_dim(input) 21 | ) 22 | return self.output 23 | 24 | def updateGradInput(self, input, gradOutput): 25 | self._backend.SoftMax_updateGradInput( 26 | self._backend.library_state, 27 | input, 28 | gradOutput, 29 | self.gradInput, 30 | self.output, 31 | self._get_dim(input) 32 | ) 33 | return self.gradInput 34 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftMin.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class SoftMin(Module): 7 | 8 | def __init__(self, dim=None): 9 | super(SoftMin, self).__init__() 10 | self.mininput = None 11 | if dim is not None: 12 | self.dim = dim 13 | 14 | def _get_dim(self, input): 15 | return getattr(self, 'dim', 0 if input.dim() == 1 or input.dim() == 3 else 1) 16 | 17 | def updateOutput(self, input): 18 | if self.mininput is None: 19 | self.mininput = input.new() 20 | self.mininput.resize_as_(input).copy_(input).mul_(-1) 21 | self._backend.SoftMax_updateOutput( 22 | self._backend.library_state, 23 | self.mininput, 24 | self.output, 25 | self._get_dim(input) 26 | ) 27 | return self.output 28 | 29 | def updateGradInput(self, input, gradOutput): 30 | if self.mininput is None: 31 | self.mininput = input.new() 32 | self.mininput.resize_as_(input).copy_(input).mul_(-1) 33 | self._backend.SoftMax_updateGradInput( 34 | self._backend.library_state, 35 | self.mininput, 36 | gradOutput, 37 | self.gradInput, 38 | self.output, 39 | self._get_dim(input) 40 | ) 41 | 42 | self.gradInput.mul_(-1) 43 | return self.gradInput 44 | 45 | def clearState(self): 46 | clear(self, 'mininput') 47 | return super(SoftMin, self).clearState() 48 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftPlus.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class SoftPlus(Module): 6 | 7 | def __init__(self, beta=1, threshold=20): 8 | super(SoftPlus, self).__init__() 9 | self.beta = beta # Beta controls sharpness of transfer function 10 | self.threshold = threshold # Avoid floating point issues with exp(x), x>20 11 | 12 | def updateOutput(self, input): 13 | # f(x) = 1/beta * log(1 + exp(beta * x)) 14 | self._backend.SoftPlus_updateOutput( 15 | self._backend.library_state, 16 | input, 17 | self.output, 18 | self.beta, 19 | self.threshold 20 | ) 21 | return self.output 22 | 23 | def updateGradInput(self, input, gradOutput): 24 | # d/dx[log(1+exp(k*x))/k] = exp(kx) / (exp(kx) + 1) 25 | # SINCE 26 | # y = (1/k)*log(1+exp(k*x)) #> x = (1/k)*log(exp(k*y)-1) 27 | # THEREFORE: 28 | # d/dx(f(x)) = (exp(k*y) - 1) / exp(k*y) 29 | self._backend.SoftPlus_updateGradInput( 30 | self._backend.library_state, 31 | input, 32 | gradOutput, 33 | self.gradInput, 34 | self.output, 35 | self.beta, 36 | self.threshold 37 | ) 38 | return self.gradInput 39 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftShrink.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class SoftShrink(Module): 6 | 7 | def __init__(self, lambd=0.5): 8 | super(SoftShrink, self).__init__() 9 | self.lambd = lambd 10 | 11 | def updateOutput(self, input): 12 | self._backend.SoftShrink_updateOutput( 13 | self._backend.library_state, 14 | input, 15 | self.output, 16 | self.lambd 17 | ) 18 | return self.output 19 | 20 | def updateGradInput(self, input, gradOutput): 21 | self._backend.SoftShrink_updateGradInput( 22 | self._backend.library_state, 23 | input, 24 | gradOutput, 25 | self.gradInput, 26 | self.lambd 27 | ) 28 | return self.gradInput 29 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SoftSign.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class SoftSign(Module): 7 | 8 | def __init__(self): 9 | super(SoftSign, self).__init__() 10 | self.temp = None 11 | self.tempgrad = None 12 | 13 | def updateOutput(self, input): 14 | if self.temp is None: 15 | self.temp = input.new() 16 | self.temp.resize_as_(input).copy_(input).abs_().add_(1) 17 | self.output.resize_as_(input).copy_(input).div_(self.temp) 18 | return self.output 19 | 20 | def updateGradInput(self, input, gradOutput): 21 | if self.tempgrad is None: 22 | self.tempgrad = input.new() 23 | self.tempgrad.resize_as_(self.output).copy_(input).abs_().add_(1).mul_(self.tempgrad) 24 | self.gradInput.resize_as_(input).copy_(gradOutput).div_(self.tempgrad) 25 | return self.gradInput 26 | 27 | def clearState(self): 28 | clear(self, 'temp', 'tempgrad') 29 | return super(SoftSign, self).clearState() 30 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SpatialAdaptiveMaxPooling.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class SpatialAdaptiveMaxPooling(Module): 7 | 8 | def __init__(self, w, h): 9 | super(SpatialAdaptiveMaxPooling, self).__init__() 10 | self.w = w 11 | self.h = h 12 | self.indices = None 13 | 14 | def updateOutput(self, input): 15 | if self.indices is None: 16 | self.indices = input.new() 17 | self.indices = self.indices.long() 18 | self._backend.SpatialAdaptiveMaxPooling_updateOutput( 19 | self._backend.library_state, 20 | input, 21 | self.output, 22 | self.indices, 23 | self.w, 24 | self.h 25 | ) 26 | return self.output 27 | 28 | def updateGradInput(self, input, gradOutput): 29 | self._backend.SpatialAdaptiveMaxPooling_updateGradInput( 30 | self._backend.library_state, 31 | input, 32 | gradOutput, 33 | self.gradInput, 34 | self.indices 35 | ) 36 | return self.gradInput 37 | 38 | def clearState(self): 39 | clear(self, 'indices') 40 | return super(SpatialAdaptiveMaxPooling, self).clearState() 41 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SpatialBatchNormalization.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .BatchNormalization import BatchNormalization 3 | 4 | 5 | class SpatialBatchNormalization(BatchNormalization): 6 | """ 7 | This class implements Batch Normalization as described in the paper: 8 | "Batch Normalization: Accelerating Deep Network Training 9 | by Reducing Internal Covariate Shift" 10 | by Sergey Ioffe, Christian Szegedy 11 | 12 | This implementation is useful for inputs coming from convolution layers. 13 | For non-convolutional layers, see BatchNormalization.lua 14 | 15 | The operation implemented is: 16 | (x - mean(x)) 17 | y = --------------------- * gamma + beta 18 | standard-deviation(x) 19 | where gamma and beta are learnable parameters. 20 | 21 | The learning of gamma and beta is optional. 22 | 23 | Usage: 24 | with learnable parameters: nn.SpatialBatchNormalization(N [, eps] [, momentum]) 25 | where N = dimensionality of input 26 | without learnable parameters: nn.SpatialBatchNormalization(N [, eps] [, momentum], False) 27 | 28 | eps is a small value added to the variance to avoid divide-by-zero. 29 | Defaults to 1e-5 30 | 31 | In training time, this layer keeps a running estimate of it's computed mean and std. 32 | The running sum is kept with a default momentum of 0.1 (unless over-ridden) 33 | In test time, this running mean/std is used to normalize. 34 | """ 35 | 36 | # expected dimension of input 37 | nDim = 4 38 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SpatialLPPooling.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .Sequential import Sequential 4 | from .Square import Square 5 | from .Power import Power 6 | from .SpatialAveragePooling import SpatialAveragePooling 7 | from .MulConstant import MulConstant 8 | from .Sqrt import Sqrt 9 | 10 | 11 | class SpatialLPPooling(Sequential): 12 | 13 | def __init__(self, nInputPlane, pnorm, kW, kH, dW=None, dH=None): 14 | super(SpatialLPPooling, self).__init__() 15 | 16 | dW = dW or kW 17 | dH = dH or kH 18 | 19 | self.kW = kW 20 | self.kH = kH 21 | self.dW = dW 22 | self.dH = dH 23 | 24 | if pnorm == 2: 25 | self.add(Square()) 26 | else: 27 | self.add(Power(pnorm)) 28 | 29 | self.add(SpatialAveragePooling(kW, kH, dW, dH)) 30 | self.add(MulConstant(kW * kH)) 31 | if pnorm == 2: 32 | self.add(Sqrt()) 33 | else: 34 | self.add(Power(1. / pnorm)) 35 | 36 | # the module is a Sequential: by default, it'll try to learn the parameters 37 | # of the sub sampler: we avoid that by redefining its methods. 38 | def reset(self, stdev=None): 39 | pass 40 | 41 | def accGradParameters(self, input, gradOutput): 42 | pass 43 | 44 | def accUpdateGradParameters(self, input, gradOutput, lr): 45 | pass 46 | 47 | def zeroGradParameters(self): 48 | pass 49 | 50 | def updateParameters(self, learningRate): 51 | pass 52 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SpatialMaxUnpooling.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .SpatialMaxPooling import SpatialMaxPooling 4 | 5 | 6 | class SpatialMaxUnpooling(Module): 7 | 8 | def __init__(self, poolingModule): 9 | super(SpatialMaxUnpooling, self).__init__() 10 | assert isinstance(poolingModule, SpatialMaxPooling) 11 | assert poolingModule.kH == poolingModule.dH 12 | assert poolingModule.kW == poolingModule.dW 13 | self.pooling = poolingModule 14 | 15 | def _setParams(self): 16 | self.indices = self.pooling.indices 17 | self.oheight = self.pooling.iheight 18 | self.owidth = self.pooling.iwidth 19 | 20 | def updateOutput(self, input): 21 | self._setParams() 22 | self._backend.SpatialMaxUnpooling_updateOutput( 23 | self._backend.library_state, 24 | input, 25 | self.output, 26 | self.indices, 27 | self.owidth, self.oheight 28 | ) 29 | return self.output 30 | 31 | def updateGradInput(self, input, gradOutput): 32 | self._setParams() 33 | self._backend.SpatialMaxUnpooling_updateGradInput( 34 | self._backend.library_state, 35 | input, 36 | gradOutput, 37 | self.gradInput, 38 | self.indices, 39 | self.owidth, self.oheight 40 | ) 41 | return self.gradInput 42 | 43 | def __repr__(self): 44 | return 'nn.SpatialMaxUnpooling associated to ' + self.pooling.__repr__() 45 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SpatialSoftMax.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class SpatialSoftMax(Module): 6 | 7 | def updateOutput(self, input): 8 | self._backend.SoftMax_updateOutput( 9 | self._backend.library_state, 10 | input, 11 | self.output, 12 | 0 if input.dim() == 1 or input.dim() == 3 else 1 13 | ) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | self._backend.SoftMax_updateGradInput( 18 | self._backend.library_state, 19 | input, 20 | gradOutput, 21 | self.gradInput, 22 | self.output, 23 | 0 if input.dim() == 1 or input.dim() == 3 else 1 24 | ) 25 | return self.gradInput 26 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/SplitTable.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class SplitTable(Module): 6 | 7 | def __init__(self, dimension): 8 | super(SplitTable, self).__init__() 9 | self.dimension = dimension 10 | 11 | def _getPositiveDimension(self, input): 12 | dimension = self.dimension 13 | if dimension < 0: 14 | dimension = input.dim() + dimension 15 | 16 | return dimension 17 | 18 | def updateOutput(self, input): 19 | dimension = self._getPositiveDimension(input) 20 | slices = input.size(dimension) 21 | 22 | currentOutput = [] 23 | for i in range(slices): 24 | currentOutput.append(input.select(dimension, i)) 25 | 26 | self.output = currentOutput 27 | return self.output 28 | 29 | def updateGradInput(self, input, gradOutput): 30 | if self.gradInput is None: 31 | return 32 | dimension = self._getPositiveDimension(input) 33 | slices = input.size(dimension) 34 | self.gradInput.resize_as_(input) 35 | 36 | for i in range(slices): 37 | self.gradInput.select(dimension, i).copy_(gradOutput[i]) 38 | 39 | return self.gradInput 40 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Sqrt.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Sqrt(Module): 6 | 7 | def __init__(self, b=0, eps=0): 8 | super(Sqrt, self).__init__() 9 | self.eps = b 10 | self.eps = eps 11 | 12 | def updateOutput(self, input): 13 | self._backend.Sqrt_updateOutput( 14 | self._backend.library_state, 15 | input, 16 | self.output, 17 | self.eps 18 | ) 19 | return self.output 20 | 21 | def updateGradInput(self, input, gradOutput): 22 | self._backend.Sqrt_updateGradInput( 23 | self._backend.library_state, 24 | input, 25 | gradOutput, 26 | self.gradInput, 27 | self.output 28 | ) 29 | return self.gradInput 30 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Square.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Square(Module): 6 | 7 | def updateOutput(self, input): 8 | self._backend.Square_updateOutput( 9 | self._backend.library_state, 10 | input, 11 | self.output 12 | ) 13 | return self.output 14 | 15 | def updateGradInput(self, input, gradOutput): 16 | self._backend.Square_updateGradInput( 17 | self._backend.library_state, 18 | input, 19 | gradOutput, 20 | self.gradInput 21 | ) 22 | return self.gradInput 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Squeeze.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Squeeze(Module): 6 | 7 | def __init__(self, dim=None): 8 | super(Squeeze, self).__init__() 9 | self.dim = dim 10 | 11 | def updateOutput(self, input): 12 | dim = self.dim 13 | self.output.set_(input.squeeze(dim) if dim is not None else input.squeeze()) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | assert input.nelement() == gradOutput.nelement() 18 | self.gradInput.set_(gradOutput.contiguous().view_as(input)) 19 | return self.gradInput 20 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Tanh.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Tanh(Module): 6 | 7 | def updateOutput(self, input): 8 | self._backend.Tanh_updateOutput( 9 | self._backend.library_state, 10 | input, 11 | self.output 12 | ) 13 | return self.output 14 | 15 | def updateGradInput(self, input, gradOutput): 16 | self._backend.Tanh_updateGradInput( 17 | self._backend.library_state, 18 | gradOutput, 19 | self.gradInput, 20 | self.output 21 | ) 22 | return self.gradInput 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/TanhShrink.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .Tanh import Tanh 4 | 5 | 6 | class TanhShrink(Module): 7 | 8 | def __init__(self): 9 | super(TanhShrink, self).__init__() 10 | self.tanh = Tanh() 11 | 12 | def updateOutput(self, input): 13 | th = self.tanh.updateOutput(input) 14 | self.output.resize_as_(input).copy_(input) 15 | self.output.add_(-1, th) 16 | return self.output 17 | 18 | def updateGradInput(self, input, gradOutput): 19 | dth = self.tanh.updateGradInput(input, gradOutput) 20 | self.gradInput.resize_as_(input).copy_(gradOutput) 21 | self.gradInput.add_(-1, dth) 22 | return self.gradInput 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/TemporalMaxPooling.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import clear 4 | 5 | 6 | class TemporalMaxPooling(Module): 7 | 8 | def __init__(self, kW, dW=None): 9 | super(TemporalMaxPooling, self).__init__() 10 | self.kW = kW 11 | self.dW = dW or kW 12 | self.indices = None 13 | 14 | def updateOutput(self, input): 15 | if self.indices is None: 16 | self.indices = input.new() 17 | self._backend.TemporalMaxPooling_updateOutput( 18 | self._backend.library_state, 19 | input, 20 | self.output, 21 | self.indices, 22 | self.kW, 23 | self.dW 24 | ) 25 | return self.output 26 | 27 | def updateGradInput(self, input, gradOutput): 28 | if self.gradInput is None: 29 | return 30 | self._backend.TemporalMaxPooling_updateGradInput( 31 | self._backend.library_state, 32 | input, 33 | gradOutput, 34 | self.gradInput, 35 | self.indices, 36 | self.kW, 37 | self.dW 38 | ) 39 | return self.gradInput 40 | 41 | def clearState(self): 42 | clear(self, 'indices') 43 | return super(TemporalMaxPooling, self).clearState() 44 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Threshold.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Threshold(Module): 6 | 7 | def __init__(self, threshold=0, value=0, inplace=False): 8 | super(Threshold, self).__init__() 9 | self.threshold = threshold 10 | self.value = value 11 | 12 | # default for inplace is False 13 | self.inplace = inplace 14 | self.validateParameters() 15 | 16 | def updateOutput(self, input): 17 | self.validateParameters() 18 | self._backend.Threshold_updateOutput( 19 | self._backend.library_state, 20 | input, 21 | self.output, 22 | self.threshold, 23 | self.value, 24 | self.inplace 25 | ) 26 | return self.output 27 | 28 | def updateGradInput(self, input, gradOutput): 29 | self.validateParameters() 30 | self._backend.Threshold_updateGradInput( 31 | self._backend.library_state, 32 | input, 33 | gradOutput, 34 | self.gradInput, 35 | self.threshold, 36 | self.value, 37 | self.inplace 38 | ) 39 | return self.gradInput 40 | 41 | def validateParameters(self): 42 | if self.inplace: 43 | if self.value > self.threshold: 44 | raise RuntimeError('in-place processing requires value ({}) to not ' 45 | 'exceed threshold ({})'.format(self.value, self.threshold)) 46 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Transpose.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class Transpose(Module): 6 | # transpose dimensions: 7 | # n = nn.Transpose({1, 4}, {1, 3}) 8 | # will transpose dims 1 and 4,: 1 and 3... 9 | 10 | def __init__(self, *args): 11 | super(Transpose, self).__init__() 12 | self.permutations = args 13 | 14 | def updateOutput(self, input): 15 | for perm in self.permutations: 16 | input = input.transpose(*perm) 17 | self.output.resize_as_(input).copy_(input) 18 | return self.output 19 | 20 | def updateGradInput(self, input, gradOutput): 21 | for perm in self.permutations[::-1]: 22 | gradOutput = gradOutput.transpose(*perm) 23 | self.gradInput.resize_as_(gradOutput).copy_(gradOutput) 24 | return self.gradInput 25 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/Unsqueeze.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .utils import addSingletondimension 4 | 5 | 6 | class Unsqueeze(Module): 7 | 8 | def __init__(self, dim): 9 | super(Unsqueeze, self).__init__() 10 | self.dim = dim 11 | 12 | def updateOutput(self, input): 13 | addSingletondimension(self.output, input, self.dim) 14 | return self.output 15 | 16 | def updateGradInput(self, input, gradOutput): 17 | assert input.nelement() == gradOutput.nelement() 18 | self.gradInput = gradOutput.contiguous().view(input.size()) 19 | return self.gradInput 20 | 21 | def __repr__(self): 22 | return super(Unsqueeze, self).__repr__() + '({})'.format(self.dim) 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/View.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | 4 | 5 | class View(Module): 6 | 7 | def resetSize(self, *args): 8 | if len(args) == 1 and isinstance(args[0], torch.Size): 9 | self.size = args[0] 10 | else: 11 | self.size = torch.Size(args) 12 | 13 | self.numElements = 1 14 | inferdim = False 15 | for i in range(len(self.size)): 16 | szi = self.size[i] 17 | if szi >= 0: 18 | self.numElements = self.numElements * self.size[i] 19 | else: 20 | assert szi == -1 21 | assert not inferdim 22 | inferdim = True 23 | 24 | return self 25 | 26 | def __init__(self, *args): 27 | super(View, self).__init__() 28 | self.resetSize(*args) 29 | 30 | def updateOutput(self, input): 31 | if self.output is None: 32 | self.output = input.new() 33 | self.output = input.view(self.size) 34 | return self.output 35 | 36 | def updateGradInput(self, input, gradOutput): 37 | if self.gradInput is None: 38 | self.gradInput = gradOutput.new() 39 | self.gradInput = gradOutput.contiguous().view(input.size()) 40 | return self.gradInput 41 | 42 | def __repr__(self): 43 | return super(View, self).__repr__() + '({})'.format(', '.join(map(str, self.size))) 44 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/nn/VolumetricBatchNormalization.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from .Module import Module 3 | from .BatchNormalization import BatchNormalization 4 | 5 | 6 | class VolumetricBatchNormalization(BatchNormalization): 7 | nDim = 5 8 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/legacy/optim/__init__.py: -------------------------------------------------------------------------------- 1 | from .adadelta import adadelta 2 | from .adagrad import adagrad 3 | from .adam import adam 4 | from .adamax import adamax 5 | from .asgd import asgd 6 | from .cg import cg 7 | from .nag import nag 8 | from .rmsprop import rmsprop 9 | from .rprop import rprop 10 | from .sgd import sgd 11 | from .lbfgs import lbfgs 12 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/multiprocessing/queue.py: -------------------------------------------------------------------------------- 1 | import io 2 | import multiprocessing 3 | import multiprocessing.queues 4 | from multiprocessing.reduction import ForkingPickler 5 | import pickle 6 | 7 | 8 | class ConnectionWrapper(object): 9 | """Proxy class for _multiprocessing.Connection which uses ForkingPickler to 10 | serialize objects""" 11 | 12 | def __init__(self, conn): 13 | self.conn = conn 14 | 15 | def send(self, obj): 16 | buf = io.BytesIO() 17 | ForkingPickler(buf, pickle.HIGHEST_PROTOCOL).dump(obj) 18 | self.send_bytes(buf.getvalue()) 19 | 20 | def recv(self): 21 | buf = self.recv_bytes() 22 | return pickle.loads(buf) 23 | 24 | def __getattr__(self, name): 25 | return getattr(self.conn, name) 26 | 27 | 28 | class Queue(multiprocessing.queues.Queue): 29 | 30 | def __init__(self, *args, **kwargs): 31 | super(Queue, self).__init__(*args, **kwargs) 32 | self._reader = ConnectionWrapper(self._reader) 33 | self._writer = ConnectionWrapper(self._writer) 34 | self._send = self._writer.send 35 | self._recv = self._reader.recv 36 | 37 | 38 | class SimpleQueue(multiprocessing.queues.SimpleQueue): 39 | 40 | def _make_methods(self): 41 | if not isinstance(self._reader, ConnectionWrapper): 42 | self._reader = ConnectionWrapper(self._reader) 43 | self._writer = ConnectionWrapper(self._writer) 44 | super(SimpleQueue, self)._make_methods() 45 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/__init__.py: -------------------------------------------------------------------------------- 1 | from .modules import * 2 | from .parameter import Parameter 3 | from .parallel import DataParallel 4 | from . import init 5 | from . import utils 6 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/_functions/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/nn/_functions/__init__.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/_functions/thnn/__init__.py: -------------------------------------------------------------------------------- 1 | _all_functions = [] 2 | 3 | from .auto import * 4 | from .normalization import * 5 | from .activation import * 6 | from .pooling import * 7 | from .sparse import * 8 | from .upsampling import * 9 | from .rnnFusedPointwise import * 10 | from .batchnorm_double_backwards import batchnorm_double_backwards_fn 11 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/_functions/thnn/auto_symbolic.py: -------------------------------------------------------------------------------- 1 | from torch.autograd._functions.utils import prepare_onnx_paddings 2 | 3 | 4 | def reflectionpad_symbolic(g, input, *params): 5 | mode = "reflect" 6 | paddings = prepare_onnx_paddings(len(input.type().sizes()), params) 7 | return g.op("Pad", input, pads_i=paddings, mode_s=mode) 8 | 9 | 10 | def replicationpad_symbolic(g, input, *params): 11 | mode = "edge" 12 | paddings = prepare_onnx_paddings(len(input.type().sizes()), params) 13 | return g.op("Pad", input, pads_i=paddings, mode_s=mode) 14 | 15 | 16 | symbolic_fns = { 17 | 'ReflectionPad1d': reflectionpad_symbolic, 18 | 'ReflectionPad2d': reflectionpad_symbolic, 19 | 'ReplicationPad1d': replicationpad_symbolic, 20 | 'ReplicationPad2d': replicationpad_symbolic, 21 | 'ReplicationPad3d': replicationpad_symbolic, 22 | } 23 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/_functions/thnn/loss.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from torch._thnn import type2backend 3 | from torch.autograd import Function 4 | 5 | from . import _all_functions 6 | from .auto import _BCELoss 7 | import warnings 8 | 9 | 10 | def _resize_weight(ctx, target): 11 | ctx.old_weight = ctx.weight 12 | if ctx.weight is not None and target.dim() != 1: 13 | ctx.weight = ctx.weight.view(1, target.size(1)).expand_as(target) 14 | 15 | 16 | def _unresize_weight(ctx): 17 | ctx.weight = ctx.old_weight 18 | del ctx.old_weight 19 | 20 | 21 | # TODO: move this code to THNN and remove _BCELoss from auto.py 22 | class BCELoss(_BCELoss): 23 | 24 | @staticmethod 25 | def forward(ctx, input, target, weight, *args): 26 | if not target.is_same_size(input): 27 | warnings.warn("Using a target size ({}) that is different to the input size ({}) is deprecated. " 28 | "Please ensure they have the same size.".format(target.size(), input.size())) 29 | assert input.nelement() == target.nelement() 30 | ctx.weight = weight 31 | _resize_weight(ctx, target) 32 | result = _BCELoss.forward(ctx, input, target, ctx.weight, *args) 33 | _unresize_weight(ctx) 34 | return result 35 | 36 | @staticmethod 37 | def backward(ctx, grad_output): 38 | target = ctx.saved_tensors[1] 39 | _resize_weight(ctx, target) 40 | result = _BCELoss.backward(ctx, grad_output) 41 | _unresize_weight(ctx) 42 | return result 43 | 44 | 45 | _all_functions.append(BCELoss) 46 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/backends/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/nn/backends/__init__.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/backends/backend.py: -------------------------------------------------------------------------------- 1 | 2 | class FunctionBackend(object): 3 | 4 | def __init__(self): 5 | self.function_classes = {} 6 | 7 | def __getattr__(self, name): 8 | fn = self.function_classes.get(name) 9 | if fn is None: 10 | raise NotImplementedError 11 | return fn 12 | 13 | def register_function(self, name, function_class): 14 | if self.function_classes.get(name): 15 | raise RuntimeError("Trying to register second function under name " + name + " in " + type(self).__name__) 16 | self.function_classes[name] = function_class 17 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/modules/normalization.py: -------------------------------------------------------------------------------- 1 | from .module import Module 2 | 3 | 4 | class CrossMapLRN2d(Module): 5 | 6 | def __init__(self, size, alpha=1e-4, beta=0.75, k=1): 7 | super(CrossMapLRN2d, self).__init__() 8 | self.size = size 9 | self.alpha = alpha 10 | self.beta = beta 11 | self.k = k 12 | 13 | def forward(self, input): 14 | return self._backend.CrossMapLRN2d(self.size, self.alpha, self.beta, 15 | self.k)(input) 16 | 17 | def __repr__(self): 18 | return self.__class__.__name__ + '(' \ 19 | + str(self.size) \ 20 | + ', alpha=' + str(self.alpha) \ 21 | + ', beta=' + str(self.beta) \ 22 | + ', k=' + str(self.k) + ')' 23 | 24 | 25 | # TODO: ContrastiveNorm2d 26 | # TODO: DivisiveNorm2d 27 | # TODO: SubtractiveNorm2d 28 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/modules/pixelshuffle.py: -------------------------------------------------------------------------------- 1 | from .module import Module 2 | from .. import functional as F 3 | 4 | 5 | class PixelShuffle(Module): 6 | r""" 7 | 对张量中形如 :math:`(*, C * r^2, H, W]` 的元素, 重新排列成 :math:`(C, H * r, W * r)`. 8 | 9 | 当使用 stride = :math:`1/r` 的高效子像素卷积很有用. 10 | 11 | 参考如下论文获得更多信息: 12 | `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network`_ 13 | Shi et. al (2016) . 14 | 15 | 参数: 16 | upscale_factor (int): 增加空间分辨率的因子 17 | 18 | 形状: 19 | - 输入: :math:`(N, C * {upscale\_factor}^2, H, W)` 20 | - 输出: :math:`(N, C, H * {upscale\_factor}, W * {upscale\_factor})` 21 | 22 | Examples:: 23 | 24 | >>> ps = nn.PixelShuffle(3) 25 | >>> input = autograd.Variable(torch.Tensor(1, 9, 4, 4)) 26 | >>> output = ps(input) 27 | >>> print(output.size()) 28 | torch.Size([1, 1, 12, 12]) 29 | 30 | .. _Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network: 31 | https://arxiv.org/abs/1609.05158 32 | """ 33 | 34 | def __init__(self, upscale_factor): 35 | super(PixelShuffle, self).__init__() 36 | self.upscale_factor = upscale_factor 37 | 38 | def forward(self, input): 39 | return F.pixel_shuffle(input, self.upscale_factor) 40 | 41 | def __repr__(self): 42 | return self.__class__.__name__ + '(upscale_factor=' + str(self.upscale_factor) + ')' 43 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/modules/utils.py: -------------------------------------------------------------------------------- 1 | import collections 2 | from itertools import repeat 3 | 4 | 5 | def _ntuple(n): 6 | def parse(x): 7 | if isinstance(x, collections.Iterable): 8 | return x 9 | return tuple(repeat(x, n)) 10 | return parse 11 | 12 | _single = _ntuple(1) 13 | _pair = _ntuple(2) 14 | _triple = _ntuple(3) 15 | _quadruple = _ntuple(4) 16 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/parallel/__init__.py: -------------------------------------------------------------------------------- 1 | from .parallel_apply import parallel_apply 2 | from .replicate import replicate 3 | from .data_parallel import DataParallel, data_parallel 4 | from .scatter_gather import scatter, gather 5 | from .distributed import DistributedDataParallel 6 | 7 | __all__ = ['replicate', 'scatter', 'parallel_apply', 'gather', 'data_parallel', 8 | 'DataParallel', 'DistributedDataParallel'] 9 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/parameter.py: -------------------------------------------------------------------------------- 1 | from torch.autograd import Variable 2 | 3 | 4 | class Parameter(Variable): 5 | r"""Variable 的一种, 常被用于 module parameter（模块参数）. 6 | 7 | Parameters 是 :class:`~torch.autograd.Variable` 的子类, 当它和 :class:`Module` 8 | 一起使用的时候会有一些特殊的属性 - 当它们被赋值给 Module 属性时, 9 | 它会自动的被加到 Module 的参数列表中, 并且会出现在 :meth:`~Module.parameters` iterator 迭代器方法中. 10 | 将 Varibale 赋值给 Module 属性则不会有这样的影响. 11 | 这样做的原因是: 我们有时候会需要缓存一些临时的 state（状态）, 12 | 例如: 模型 RNN 中的最后一个隐藏状态. 13 | 如果没有 :class:`Parameter` 这个类的话, 14 | 那么这些临时表也会注册为模型变量. 15 | 16 | Variable 与 Parameter 的另一个不同之处在于, 17 | Parameter 不能被 volatile (即: 无法设置 volatile=True) 而且默认 requires_grad=True. 18 | Variable 默认 requires_grad=False. 19 | 20 | Arguments: 21 | data (Tensor): parameter tensor. 22 | requires_grad (bool, optional): 如果参数需要梯度. 更多细节请参阅 :ref:`excluding-subgraphs`. 23 | """ 24 | def __new__(cls, data=None, requires_grad=True): 25 | return super(Parameter, cls).__new__(cls, data, requires_grad=requires_grad) 26 | 27 | def __repr__(self): 28 | return 'Parameter containing:' + self.data.__repr__() 29 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/utils/__init__.py: -------------------------------------------------------------------------------- 1 | from . import rnn 2 | from .clip_grad import clip_grad_norm 3 | from .weight_norm import weight_norm, remove_weight_norm 4 | from .convert_parameters import parameters_to_vector, vector_to_parameters 5 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/nn/utils/clip_grad.py: -------------------------------------------------------------------------------- 1 | 2 | def clip_grad_norm(parameters, max_norm, norm_type=2): 3 | r"""接收一个包含 Variable 的可迭代对象, 对 Variable 的梯度按范数进行裁剪. 4 | 5 | 范数是对所有梯度进行计算的, 等价于把所有输入变量的梯度连接成一个向量, 然后对这个向量按范数进行裁剪. 梯度将会被原地修改. 6 | 7 | Arguments: 8 | parameters (Iterable[Variable]): 一个可迭代对象, 其包含将要进行梯度正规化的 Variable 9 | max_norm (float or int): 梯度的最大范数 10 | norm_type (float or int): p 范数(指定 p ). 用 ``'inf'`` 表示无穷范数 11 | 12 | Returns: 13 | 梯度的范数 (视为单个向量的). 14 | """ 15 | parameters = list(filter(lambda p: p.grad is not None, parameters)) 16 | max_norm = float(max_norm) 17 | norm_type = float(norm_type) 18 | if norm_type == float('inf'): 19 | total_norm = max(p.grad.data.abs().max() for p in parameters) 20 | else: 21 | total_norm = 0 22 | for p in parameters: 23 | param_norm = p.grad.data.norm(norm_type) 24 | total_norm += param_norm ** norm_type 25 | total_norm = total_norm ** (1. / norm_type) 26 | clip_coef = max_norm / (total_norm + 1e-6) 27 | if clip_coef < 1: 28 | for p in parameters: 29 | p.grad.data.mul_(clip_coef) 30 | return total_norm 31 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/optim/__init__.py: -------------------------------------------------------------------------------- 1 | """ 2 | :mod:`torch.optim` is a package implementing various optimization algorithms. 3 | Most commonly used methods are already supported, and the interface is general 4 | enough, so that more sophisticated ones can be also easily integrated in the 5 | future. 6 | """ 7 | 8 | from .adadelta import Adadelta 9 | from .adagrad import Adagrad 10 | from .adam import Adam 11 | from .sparse_adam import SparseAdam 12 | from .adamax import Adamax 13 | from .asgd import ASGD 14 | from .sgd import SGD 15 | from .rprop import Rprop 16 | from .rmsprop import RMSprop 17 | from .optimizer import Optimizer 18 | from .lbfgs import LBFGS 19 | from . import lr_scheduler 20 | 21 | del adadelta 22 | del adagrad 23 | del adam 24 | del adamax 25 | del asgd 26 | del sgd 27 | del rprop 28 | del rmsprop 29 | del optimizer 30 | del lbfgs 31 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/doc/zh/source/code/torch/utils/__init__.py -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/backcompat/__init__.py: -------------------------------------------------------------------------------- 1 | from torch._C import _set_backcompat_broadcast_warn 2 | from torch._C import _get_backcompat_broadcast_warn 3 | from torch._C import _set_backcompat_keepdim_warn 4 | from torch._C import _get_backcompat_keepdim_warn 5 | 6 | 7 | class Warning(object): 8 | def __init__(self, setter, getter): 9 | self.setter = setter 10 | self.getter = getter 11 | 12 | def set_enabled(self, value): 13 | self.setter(value) 14 | 15 | def get_enabled(self): 16 | return self.getter() 17 | 18 | enabled = property(get_enabled, set_enabled) 19 | 20 | broadcast_warning = Warning(_set_backcompat_broadcast_warn, _get_backcompat_broadcast_warn) 21 | keepdim_warning = Warning(_set_backcompat_keepdim_warn, _get_backcompat_keepdim_warn) 22 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/data/__init__.py: -------------------------------------------------------------------------------- 1 | 2 | from .dataset import Dataset, TensorDataset, ConcatDataset 3 | from .dataloader import DataLoader 4 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/dlpack.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | from torch._C import _from_dlpack as from_dlpack 4 | from torch._C import _to_dlpack as to_dlpack 5 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/hooks.py: -------------------------------------------------------------------------------- 1 | import collections 2 | import weakref 3 | 4 | 5 | class RemovableHandle(object): 6 | """A handle which provides the capability to remove a hook.""" 7 | 8 | next_id = 0 9 | 10 | def __init__(self, hooks_dict): 11 | self.hooks_dict_ref = weakref.ref(hooks_dict) 12 | self.id = RemovableHandle.next_id 13 | RemovableHandle.next_id += 1 14 | 15 | def remove(self): 16 | hooks_dict = self.hooks_dict_ref() 17 | if hooks_dict is not None and self.id in hooks_dict: 18 | del hooks_dict[self.id] 19 | 20 | def __getstate__(self): 21 | return (self.hooks_dict_ref(), self.id) 22 | 23 | def __setstate__(self, state): 24 | if state[0] is None: 25 | # create a dead reference 26 | self.hooks_dict_ref = weakref.ref(collections.OrderedDict()) 27 | else: 28 | self.hooks_dict_ref = weakref.ref(state[0]) 29 | self.id = state[1] 30 | RemovableHandle.next_id = max(RemovableHandle.next_id, self.id + 1) 31 | 32 | def __enter__(self): 33 | return self 34 | 35 | def __exit__(self, type, value, tb): 36 | self.remove() 37 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/serialization/__init__.py: -------------------------------------------------------------------------------- 1 | 2 | from .read_lua_file import load_lua, T7Reader 3 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/__init__.py: -------------------------------------------------------------------------------- 1 | 2 | from .trainer import Trainer 3 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/__init__.py: -------------------------------------------------------------------------------- 1 | from .progress import ProgressMonitor 2 | from .accuracy import AccuracyMonitor 3 | from .time import TimeMonitor 4 | from .loss import LossMonitor 5 | from .logger import Logger 6 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/accuracy.py: -------------------------------------------------------------------------------- 1 | from .monitor import Monitor 2 | 3 | 4 | class AccuracyMonitor(Monitor): 5 | stat_name = 'accuracy' 6 | 7 | def __init__(self, *args, **kwargs): 8 | kwargs.setdefault('unit', '%') 9 | kwargs.setdefault('precision', 2) 10 | super(AccuracyMonitor, self).__init__(*args, **kwargs) 11 | 12 | def _get_value(self, iteration, input, target, output, loss): 13 | batch_size = input.size(0) 14 | predictions = output.max(1)[1].type_as(target) 15 | correct = predictions.eq(target) 16 | if not hasattr(correct, 'sum'): 17 | correct = correct.cpu() 18 | correct = correct.sum() 19 | return 100. * correct / batch_size 20 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/loss.py: -------------------------------------------------------------------------------- 1 | from .monitor import Monitor 2 | 3 | 4 | class LossMonitor(Monitor): 5 | stat_name = 'loss' 6 | 7 | def _get_value(self, iteration, input, target, output, loss): 8 | return loss[0] 9 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/plugin.py: -------------------------------------------------------------------------------- 1 | 2 | class Plugin(object): 3 | 4 | def __init__(self, interval=None): 5 | if interval is None: 6 | interval = [] 7 | self.trigger_interval = interval 8 | 9 | def register(self, trainer): 10 | raise NotImplementedError 11 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/progress.py: -------------------------------------------------------------------------------- 1 | from .plugin import Plugin 2 | 3 | 4 | class ProgressMonitor(Plugin): 5 | stat_name = 'progress' 6 | 7 | def __init__(self): 8 | super(ProgressMonitor, self).__init__([(1, 'iteration'), (1, 'epoch')]) 9 | 10 | def register(self, trainer): 11 | self.trainer = trainer 12 | stats = self.trainer.stats.setdefault(self.stat_name, {}) 13 | stats['samples_used'] = 0 14 | stats['epoch_size'] = len(trainer.dataset) 15 | stats['log_iter_fields'] = [ 16 | '{samples_used}/{epoch_size}', 17 | '({percent:.2f}%)' 18 | ] 19 | 20 | def iteration(self, iteration, input, *args): 21 | stats = self.trainer.stats.setdefault(self.stat_name, {}) 22 | stats['samples_used'] += 1 23 | stats['percent'] = 100. * stats['samples_used'] / stats['epoch_size'] 24 | 25 | def epoch(self, *args): 26 | stats = self.trainer.stats.setdefault(self.stat_name, {}) 27 | stats['samples_used'] = 0 28 | stats['percent'] = 0 29 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/utils/trainer/plugins/time.py: -------------------------------------------------------------------------------- 1 | from __future__ import absolute_import 2 | import time 3 | 4 | from .monitor import Monitor 5 | 6 | 7 | class TimeMonitor(Monitor): 8 | stat_name = 'time' 9 | 10 | def __init__(self, *args, **kwargs): 11 | kwargs.setdefault('unit', 'ms') 12 | kwargs.setdefault('precision', 0) 13 | super(TimeMonitor, self).__init__(*args, **kwargs) 14 | self.last_time = None 15 | 16 | def _get_value(self, *args): 17 | if self.last_time: 18 | now = time.time() 19 | duration = now - self.last_time 20 | self.last_time = now 21 | return duration * 1000 22 | else: 23 | self.last_time = time.time() 24 | return 0 25 | -------------------------------------------------------------------------------- /doc/zh/source/code/torch/version.py: -------------------------------------------------------------------------------- 1 | __version__ = '0.3.0.post4' 2 | debug = False 3 | cuda = '8.0.61' 4 | -------------------------------------------------------------------------------- /doc/zh/source/code/torchvision/__init__.py: -------------------------------------------------------------------------------- 1 | from torchvision import models 2 | from torchvision import datasets 3 | from torchvision import transforms 4 | from torchvision import utils 5 | 6 | __version__ = '0.2.0' 7 | 8 | _image_backend = 'PIL' 9 | 10 | 11 | def set_image_backend(backend): 12 | """ 13 | 指定用于加载图像的包。 14 | 15 | Args: 16 | backend (string): 图像处理后端的名称. {'PIL', 'accimage'}之一. 17 | :mod:`accimage` 使用 Intel IPP library（高性能图像加载和增强程序模拟的程序）. 18 | 通常比PIL库（）要快, 但是不支持许多操作. 19 | """ 20 | global _image_backend 21 | if backend not in ['PIL', 'accimage']: 22 | raise ValueError("Invalid backend '{}'. Options are 'PIL' and 'accimage'" 23 | .format(backend)) 24 | _image_backend = backend 25 | 26 | 27 | def get_image_backend(): 28 | """ 29 | 获取用于加载图像的包的名称 30 | """ 31 | return _image_backend 32 | -------------------------------------------------------------------------------- /doc/zh/source/code/torchvision/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | from .lsun import LSUN, LSUNClass 2 | from .folder import ImageFolder 3 | from .coco import CocoCaptions, CocoDetection 4 | from .cifar import CIFAR10, CIFAR100 5 | from .stl10 import STL10 6 | from .mnist import MNIST, FashionMNIST 7 | from .svhn import SVHN 8 | from .phototour import PhotoTour 9 | from .fakedata import FakeData 10 | from .semeion import SEMEION 11 | 12 | __all__ = ('LSUN', 'LSUNClass', 13 | 'ImageFolder', 'FakeData', 14 | 'CocoCaptions', 'CocoDetection', 15 | 'CIFAR10', 'CIFAR100', 'FashionMNIST', 16 | 'MNIST', 'STL10', 'SVHN', 'PhotoTour', 'SEMEION') 17 | -------------------------------------------------------------------------------- /doc/zh/source/code/torchvision/datasets/utils.py: -------------------------------------------------------------------------------- 1 | import os 2 | import os.path 3 | import hashlib 4 | import errno 5 | 6 | 7 | def check_integrity(fpath, md5): 8 | if not os.path.isfile(fpath): 9 | return False 10 | md5o = hashlib.md5() 11 | with open(fpath, 'rb') as f: 12 | # read in 1MB chunks 13 | for chunk in iter(lambda: f.read(1024 * 1024), b''): 14 | md5o.update(chunk) 15 | md5c = md5o.hexdigest() 16 | if md5c != md5: 17 | return False 18 | return True 19 | 20 | 21 | def download_url(url, root, filename, md5): 22 | from six.moves import urllib 23 | 24 | root = os.path.expanduser(root) 25 | fpath = os.path.join(root, filename) 26 | 27 | try: 28 | os.makedirs(root) 29 | except OSError as e: 30 | if e.errno == errno.EEXIST: 31 | pass 32 | else: 33 | raise 34 | 35 | # downloads file 36 | if os.path.isfile(fpath) and check_integrity(fpath, md5): 37 | print('Using downloaded and verified file: ' + fpath) 38 | else: 39 | try: 40 | print('Downloading ' + url + ' to ' + fpath) 41 | urllib.request.urlretrieve(url, fpath) 42 | except: 43 | if url[:5] == 'https': 44 | url = url.replace('https:', 'http:') 45 | print('Failed download. Trying https -> http instead.' 46 | ' Downloading ' + url + ' to ' + fpath) 47 | urllib.request.urlretrieve(url, fpath) 48 | -------------------------------------------------------------------------------- /doc/zh/source/code/torchvision/models/__init__.py: -------------------------------------------------------------------------------- 1 | from .alexnet import * 2 | from .resnet import * 3 | from .vgg import * 4 | from .squeezenet import * 5 | from .inception import * 6 | from .densenet import * 7 | -------------------------------------------------------------------------------- /doc/zh/source/code/torchvision/transforms/__init__.py: -------------------------------------------------------------------------------- 1 | from .transforms import * 2 | -------------------------------------------------------------------------------- /doc/zh/source/cuda.rst: -------------------------------------------------------------------------------- 1 | torch.cuda 2 | =================================== 3 | 4 | .. currentmodule:: torch.cuda 5 | 6 | .. automodule:: torch.cuda 7 | :members: 8 | 9 | Random Number Generator 10 | ------------------------- 11 | .. autofunction:: get_rng_state 12 | .. autofunction:: set_rng_state 13 | .. autofunction:: manual_seed 14 | .. autofunction:: manual_seed_all 15 | .. autofunction:: seed 16 | .. autofunction:: seed_all 17 | .. autofunction:: initial_seed 18 | 19 | 20 | Communication collectives 21 | ------------------------- 22 | 23 | .. autofunction:: torch.cuda.comm.broadcast 24 | 25 | .. autofunction:: torch.cuda.comm.reduce_add 26 | 27 | .. autofunction:: torch.cuda.comm.scatter 28 | 29 | .. autofunction:: torch.cuda.comm.gather 30 | 31 | Streams and events 32 | ------------------ 33 | 34 | .. autoclass:: Stream 35 | :members: 36 | 37 | .. autoclass:: Event 38 | :members: 39 | 40 | Memory management 41 | ----------------- 42 | .. autofunction:: empty_cache 43 | 44 | NVIDIA Tools Extension (NVTX) 45 | ----------------------------- 46 | 47 | .. autofunction:: torch.cuda.nvtx.mark 48 | .. autofunction:: torch.cuda.nvtx.range_push 49 | .. autofunction:: torch.cuda.nvtx.range_pop 50 | -------------------------------------------------------------------------------- /doc/zh/source/data.rst: -------------------------------------------------------------------------------- 1 | torch.utils.data 2 | =================================== 3 | 4 | .. automodule:: torch.utils.data 5 | .. autoclass:: Dataset 6 | .. autoclass:: TensorDataset 7 | .. autoclass:: ConcatDataset 8 | .. autoclass:: DataLoader 9 | .. autoclass:: torch.utils.data.sampler.Sampler 10 | .. autoclass:: torch.utils.data.sampler.SequentialSampler 11 | .. autoclass:: torch.utils.data.sampler.RandomSampler 12 | .. autoclass:: torch.utils.data.sampler.SubsetRandomSampler 13 | .. autoclass:: torch.utils.data.sampler.WeightedRandomSampler 14 | .. autoclass:: torch.utils.data.distributed.DistributedSampler 15 | -------------------------------------------------------------------------------- /doc/zh/source/distributions.rst: -------------------------------------------------------------------------------- 1 | .. role:: hidden 2 | :class: hidden-section 3 | 4 | Probability distributions - torch.distributions 5 | ================================================== 6 | 7 | .. automodule:: torch.distributions 8 | .. currentmodule:: torch.distributions 9 | 10 | Distribution（概率分布） 11 | ~~~~~~~~~~~~~~~~~~~~~~~ 12 | 13 | .. autoclass:: Distribution 14 | :members: 15 | 16 | Bernoulli（伯努利分布） 17 | ~~~~~~~~~~~~~~~~~~~~~~~ 18 | 19 | .. autoclass:: Bernoulli 20 | :members: 21 | 22 | Categorical（类别分布） 23 | ~~~~~~~~~~~~~~~~~~~~~~~ 24 | 25 | .. autoclass:: Categorical 26 | :members: 27 | 28 | Normal（正态分布） 29 | ~~~~~~~~~~~~~~~~~~~~~~~ 30 | 31 | .. autoclass:: Normal 32 | :members: 33 | -------------------------------------------------------------------------------- /doc/zh/source/ffi.rst: -------------------------------------------------------------------------------- 1 | torch.utils.ffi 2 | =============== 3 | 4 | .. currentmodule:: torch.utils.ffi 5 | .. autofunction:: create_extension 6 | 7 | -------------------------------------------------------------------------------- /doc/zh/source/index.rst: -------------------------------------------------------------------------------- 1 | .. PyTorch documentation master file, created by 2 | sphinx-quickstart on Fri Dec 23 13:31:47 2016. 3 | You can adapt this file completely to your liking, but it should at least 4 | contain the root `toctree` directive. 5 | 6 | :github_url: https://github.com/apachecn/pytorch-doc-zh 7 | 8 | PyTorch 0.3.0 中文文档 9 | =================================== 10 | 11 | PyTorch 是一个针对 deep learning（深度学习）, 并且使用 GPU 和 CPU 来优化的 tensor library（张量库）. 12 | 13 | .. toctree:: 14 | :glob: 15 | :maxdepth: 1 16 | :caption: 介绍 17 | 18 | notes/* 19 | 20 | 21 | .. toctree:: 22 | :maxdepth: 1 23 | :caption: Package 参考 24 | 25 | torch 26 | tensors 27 | sparse 28 | storage 29 | nn 30 | optim 31 | torch.autograd 32 | torch.distributions 33 | torch.multiprocessing 34 | torch.distributed 35 | torch.legacy 36 | cuda 37 | ffi 38 | data 39 | model_zoo 40 | onnx 41 | 42 | .. toctree:: 43 | :glob: 44 | :maxdepth: 2 45 | :caption: torchvision 参考 46 | 47 | torchvision/index 48 | 49 | .. toctree:: 50 | :maxdepth: 1 51 | :caption: 项目相关 52 | 53 | 项目贡献者 54 | 组织学习交流群 55 | 56 | 57 | 58 | 索引表 59 | ================== 60 | 61 | * :ref:`genindex` 62 | * :ref:`modindex` 63 | 64 | 相关链接 65 | ================== 66 | - GitHub: `apachecn/pytorch-doc-zh `__ 67 | - 项目贡献者: ``__ 68 | - 组织学习交流群: ``__ -------------------------------------------------------------------------------- /doc/zh/source/legacy.rst: -------------------------------------------------------------------------------- 1 | Legacy package - torch.legacy 2 | =================================== 3 | 4 | .. automodule:: torch.legacy 5 | -------------------------------------------------------------------------------- /doc/zh/source/model_zoo.rst: -------------------------------------------------------------------------------- 1 | torch.utils.model_zoo 2 | =================================== 3 | 4 | .. automodule:: torch.utils.model_zoo 5 | .. autofunction:: load_url 6 | -------------------------------------------------------------------------------- /doc/zh/source/notes/serialization.rst: -------------------------------------------------------------------------------- 1 | 2 | 序列化语义 3 | ======================= 4 | 5 | 最佳实践 6 | -------------- 7 | 8 | .. _recommend-saving-models: 9 | 10 | 保存模型的推荐方法 11 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 12 | 13 | 有两种主要的方法可以用来序列化和恢复模型. 14 | 15 | 第一种方法（推荐）, 只保存和加载模型的参数:: 16 | 17 | torch.save(the_model.state_dict(), PATH) 18 | 19 | 然后:: 20 | 21 | the_model = TheModelClass(*args, **kwargs) 22 | the_model.load_state_dict(torch.load(PATH)) 23 | 24 | 第二种方法, 保存和加载整个模型:: 25 | 26 | torch.save(the_model, PATH) 27 | 28 | 然后:: 29 | 30 | the_model = torch.load(PATH) 31 | 32 | 但是在这种情况下, 序列化的数据与特定的类和固定的目录结构绑定, 所以当它被用于其他项目中, 或者经过一些重大的重构之后, 可能会以各种各样的方式崩掉. 33 | -------------------------------------------------------------------------------- /doc/zh/source/storage.rst: -------------------------------------------------------------------------------- 1 | torch.Storage 2 | =================================== 3 | 4 | 一个 :class:`torch.Storage` 是一个连续的一维的单一数据类型的数组. 5 | 6 | 每个 :class:`torch.Tensor` 都有一个对应的相同数据类型的存储. 7 | 8 | .. autoclass:: torch.FloatStorage 9 | :members: 10 | :undoc-members: 11 | :inherited-members: 12 | -------------------------------------------------------------------------------- /doc/zh/source/torchvision/index.rst: -------------------------------------------------------------------------------- 1 | torchvision 2 | =========== 3 | 4 | 模块 :mod:`torchvision` 库包含了计算机视觉中一些常用的数据集, 5 | 模型架构以及图像变换方法. 6 | 7 | .. toctree:: 8 | :maxdepth: 2 9 | :caption: Package Reference 10 | 11 | datasets 12 | models 13 | transforms 14 | utils 15 | 16 | .. automodule:: torchvision 17 | :members: -------------------------------------------------------------------------------- /doc/zh/source/torchvision/transforms.rst: -------------------------------------------------------------------------------- 1 | torchvision.transforms 2 | ====================== 3 | 4 | .. currentmodule:: torchvision.transforms 5 | 6 | Transforms（变换）是常见的 image transforms（图像变换）.他们可以使用 :class:`Compose` 类以链在一起来进行操作. 7 | 8 | .. autoclass:: Compose 9 | 10 | Transforms PIL Image 上的变换 11 | ----------------------------- 12 | 13 | .. autoclass:: Resize 14 | 15 | .. autoclass:: Scale 16 | 17 | .. autoclass:: CenterCrop 18 | 19 | .. autoclass:: RandomCrop 20 | 21 | .. autoclass:: RandomHorizontalFlip 22 | 23 | .. autoclass:: RandomVerticalFlip 24 | 25 | .. autoclass:: RandomResizedCrop 26 | 27 | .. autoclass:: RandomSizedCrop 28 | 29 | .. autoclass:: Grayscale 30 | 31 | .. autoclass:: RandomGrayscale 32 | 33 | .. autoclass:: FiveCrop 34 | 35 | .. autoclass:: TenCrop 36 | 37 | .. autoclass:: Pad 38 | 39 | .. autoclass:: ColorJitter 40 | 41 | torch.\*Tensor 上的变换 42 | ---------------------------- 43 | 44 | .. autoclass:: Normalize 45 | :members: __call__ 46 | :special-members: 47 | 48 | 49 | 转换类型的变换 50 | --------------------- 51 | 52 | .. autoclass:: ToTensor 53 | :members: __call__ 54 | :special-members: 55 | 56 | .. autoclass:: ToPILImage 57 | :members: __call__ 58 | :special-members: 59 | 60 | 通用的变换 61 | ------------------ 62 | 63 | .. autoclass:: Lambda -------------------------------------------------------------------------------- /doc/zh/source/torchvision/utils.rst: -------------------------------------------------------------------------------- 1 | torchvision.utils 2 | ================= 3 | 4 | .. currentmodule:: torchvision.utils 5 | 6 | .. autofunction:: make_grid 7 | 8 | .. autofunction:: save_image -------------------------------------------------------------------------------- /tutorial/en/LICENSE: -------------------------------------------------------------------------------- 1 | BSD 3-Clause License 2 | 3 | Copyright (c) 2017, Pytorch contributors 4 | All rights reserved. 5 | 6 | Redistribution and use in source and binary forms, with or without 7 | modification, are permitted provided that the following conditions are met: 8 | 9 | * Redistributions of source code must retain the above copyright notice, this 10 | list of conditions and the following disclaimer. 11 | 12 | * Redistributions in binary form must reproduce the above copyright notice, 13 | this list of conditions and the following disclaimer in the documentation 14 | and/or other materials provided with the distribution. 15 | 16 | * Neither the name of the copyright holder nor the names of its 17 | contributors may be used to endorse or promote products derived from 18 | this software without specific prior written permission. 19 | 20 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 21 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 23 | DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE 24 | FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 26 | SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 27 | CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 28 | OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 29 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 | -------------------------------------------------------------------------------- /tutorial/en/_static/img/SRResNet.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/SRResNet.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/Variable.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/Variable.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/cartpole.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/cartpole.gif -------------------------------------------------------------------------------- /tutorial/en/_static/img/cat.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/cat.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/cat_224x224.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/cat_224x224.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/cat_output1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/cat_output1.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/char_rnn_generation.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/char_rnn_generation.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/cifar10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/cifar10.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/data_parallel.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/data_parallel.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/DistPyTorch.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/DistPyTorch.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/all_gather.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/all_gather.pdf -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/all_gather.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/all_gather.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/all_reduce.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/all_reduce.pdf -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/all_reduce.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/all_reduce.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/broadcast.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/broadcast.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/gather.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/gather.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/reduce.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/reduce.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/scatter.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/scatter.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/send_recv.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/send_recv.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/distributed/send_recv_big.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/distributed/send_recv_big.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/dynamic_graph.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/dynamic_graph.gif -------------------------------------------------------------------------------- /tutorial/en/_static/img/landmarked_face2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/landmarked_face2.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/mnist.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/mnist.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/neural-style/dancing.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/neural-style/dancing.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/neural-style/neuralstyle.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/neural-style/neuralstyle.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/neural-style/picasso.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/neural-style/picasso.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/pytorch-logo-dark.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/pytorch-logo-dark.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/attention-decoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden;encoder_outputs} 11 | 12 | 13 | input -> embedding; 14 | embedding -> dropout; 15 | dropout -> embedded; 16 | 17 | embedded -> attn; 18 | prev_hidden -> attn; 19 | attn -> attn_softmax; 20 | attn_softmax -> attn_weights; 21 | attn_weights -> bmm; 22 | encoder_outputs -> bmm; 23 | bmm -> attn_applied; 24 | attn_applied -> attn_combine; 25 | embedded -> attn_combine; 26 | 27 | attn_combine -> relu -> gru; 28 | prev_hidden -> gru; 29 | gru -> out; 30 | gru -> hidden; 31 | 32 | out -> softmax; 33 | softmax -> output; 34 | 35 | {rank=same;output;hidden} 36 | 37 | // Layer nodes 38 | embedding [fillcolor=dodgerblue, fontcolor=white]; 39 | attn [fillcolor=dodgerblue, fontcolor=white]; 40 | attn_combine [fillcolor=dodgerblue, fontcolor=white]; 41 | bmm [fillcolor=dodgerblue, fontcolor=white]; 42 | gru [fillcolor=dodgerblue, fontcolor=white]; 43 | out [fillcolor=dodgerblue, fontcolor=white]; 44 | 45 | // Function nodes 46 | dropout [fillcolor=palegreen]; 47 | relu [fillcolor=palegreen]; 48 | softmax [fillcolor=palegreen]; 49 | attn_softmax [fillcolor=palegreen]; 50 | 51 | } 52 | -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/attention-decoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/attention-decoder-network.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/decoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden} 11 | 12 | input -> embedding; 13 | embedding -> relu; 14 | relu -> gru; 15 | 16 | prev_hidden -> gru; 17 | gru -> out; 18 | gru -> hidden; 19 | 20 | out -> softmax; 21 | softmax -> output; 22 | 23 | {rank=same;output;hidden} 24 | 25 | // Layer nodes 26 | embedding [fillcolor=dodgerblue, fontcolor=white]; 27 | gru [fillcolor=dodgerblue, fontcolor=white]; 28 | out [fillcolor=dodgerblue, fontcolor=white]; 29 | 30 | // Function nodes 31 | relu [fillcolor=palegreen]; 32 | softmax [fillcolor=palegreen]; 33 | 34 | } 35 | -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/decoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/decoder-network.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/decoder.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/decoder.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/decoder@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/decoder@2x.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/encoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden} 11 | 12 | input -> embedding; 13 | embedding -> embedded; 14 | embedded -> gru; 15 | prev_hidden -> gru; 16 | gru -> output; 17 | gru -> hidden; 18 | 19 | embedding [fillcolor=dodgerblue, fontcolor=white]; 20 | gru [fillcolor=dodgerblue, fontcolor=white]; 21 | 22 | } 23 | -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/encoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/encoder-network.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/seq2seq.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/seq2seq.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/seq2seq@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/seq2seq@2x.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/word-encoding.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/word-encoding.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq-seq-images/word-encoding@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq-seq-images/word-encoding@2x.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/seq2seq_flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/seq2seq_flat.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/stn/FSeq.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/stn/FSeq.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/stn/Five.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/stn/Five.gif -------------------------------------------------------------------------------- /tutorial/en/_static/img/stn/stn-arch.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/stn/stn-arch.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/stn/tr.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/stn/tr.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/tensor_illustration.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/tensor_illustration.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/tensor_illustration_flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/tensor_illustration_flat.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/thumbnails/babel.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/thumbnails/babel.jpg -------------------------------------------------------------------------------- /tutorial/en/_static/img/thumbnails/default.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/thumbnails/default.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/thumbnails/examples.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/thumbnails/examples.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/thumbnails/pytorch-logo-flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/thumbnails/pytorch-logo-flat.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/thumbnails/torch-logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/thumbnails/torch-logo.png -------------------------------------------------------------------------------- /tutorial/en/_static/img/torch-nn-vs-pytorch-nn.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/_static/img/torch-nn-vs-pytorch-nn.png -------------------------------------------------------------------------------- /tutorial/en/_templates/layout.html: -------------------------------------------------------------------------------- 1 | {% extends "!layout.html" %} 2 | 3 | {% block footer %} 4 | {{ super() }} 5 | 6 | 13 | 22 | 36 | {% endblock %} -------------------------------------------------------------------------------- /tutorial/en/advanced_source/README.txt: -------------------------------------------------------------------------------- 1 | Advanced Tutorials 2 | ------------------ 3 | 4 | 1. neural_style_tutorial.py 5 | Neural Transfer with PyTorch 6 | http://pytorch.org/tutorials/advanced/neural_style_tutorial.html 7 | 8 | 2. numpy_extensions_tutorial.py 9 | Creating extensions using numpy and scipy 10 | http://pytorch.org/tutorials/advanced/numpy_extensions_tutorial.html 11 | 12 | 3. c_extension.rst 13 | Custom C extensions for pytorch 14 | http://pytorch.org/tutorials/advanced/c_extension.html 15 | 16 | 4. super_resolution_with_caffe2.py 17 | Transfering a model from PyTorch to Caffe2 and Mobile using ONNX 18 | http://pytorch.org/tutorials/advanced/super_resolution_with_caffe2.html 19 | -------------------------------------------------------------------------------- /tutorial/en/beginner_source/README.txt: -------------------------------------------------------------------------------- 1 | Beginner Tutorials 2 | ------------------ 3 | 4 | 1. blitz/* and deep_learning_60min_blitz.rst 5 | Deep Learning with PyTorch: A 60 Minute Blitz 6 | http://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html 7 | 8 | 2. former_torches/* and former_torchies_tutorial.rst 9 | PyTorch for former Torch users 10 | http://pytorch.org/tutorials/beginner/former_torchies_tutorial.html 11 | 12 | 3. examples_*/* and pytorch_with_examples.rst 13 | Learning PyTorch with Examples 14 | http://pytorch.org/tutorials/beginner/pytorch_with_examples.html 15 | 16 | 4. transfer_learning_tutorial.py 17 | Transfer Learning tutorial 18 | http://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html 19 | 20 | 5. nlp/* and deep_learning_nlp_tutorial.rst 21 | Deep Learning for NLP with Pytorch 22 | http://pytorch.org/tutorials/beginner/deep_learning_nlp_tutorial.html -------------------------------------------------------------------------------- /tutorial/en/beginner_source/blitz/README.txt: -------------------------------------------------------------------------------- 1 | Deep Learning with PyTorch: A 60 Minute Blitz 2 | --------------------------------------------- 3 | 4 | 1. tensor_tutorial.py 5 | What is PyTorch? 6 | http://pytorch.org/tutorials/beginner/blitz/tensor_tutorial.html 7 | 8 | 2. autograd_tutorial.py 9 | Autograd: automatic differentiation 10 | http://pytorch.org/tutorials/beginner/blitz/autograd_tutorial.html 11 | 12 | 3. neural_networks_tutorial.py 13 | Neural Networks 14 | http://pytorch.org/tutorials/beginner/blitz/neural_networks_tutorial.html# 15 | 16 | 4. cifar10_tutorial.py 17 | Training a classifier 18 | http://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html -------------------------------------------------------------------------------- /tutorial/en/beginner_source/deep_learning_60min_blitz.rst: -------------------------------------------------------------------------------- 1 | Deep Learning with PyTorch: A 60 Minute Blitz 2 | --------------------------------------------- 3 | **Author**: `Soumith Chintala `_ 4 | 5 | Goal of this tutorial: 6 | 7 | - Understand PyTorch’s Tensor library and neural networks at a high 8 | level. 9 | - Train a small neural network to classify images 10 | 11 | *This tutorial assumes that you have a basic familiarity of numpy* 12 | 13 | .. Note:: 14 | Make sure you have the `torch`_ and `torchvision`_ packages installed. 15 | 16 | .. _torch: https://github.com/pytorch/pytorch 17 | .. _torchvision: https://github.com/pytorch/vision 18 | 19 | 20 | .. toctree:: 21 | :hidden: 22 | 23 | /beginner/blitz/tensor_tutorial 24 | /beginner/blitz/autograd_tutorial 25 | /beginner/blitz/neural_networks_tutorial 26 | /beginner/blitz/cifar10_tutorial 27 | /beginner/blitz/data_parallel_tutorial 28 | 29 | .. galleryitem:: /beginner/blitz/tensor_tutorial.py 30 | :figure: /_static/img/tensor_illustration_flat.png 31 | 32 | .. galleryitem:: /beginner/blitz/autograd_tutorial.py 33 | :figure: /_static/img/Variable.png 34 | 35 | .. galleryitem:: /beginner/blitz/neural_networks_tutorial.py 36 | :figure: /_static/img/mnist.png 37 | 38 | .. galleryitem:: /beginner/blitz/cifar10_tutorial.py 39 | :figure: /_static/img/cifar10.png 40 | 41 | .. galleryitem:: /beginner/blitz/data_parallel_tutorial.py 42 | :figure: /_static/img/data_parallel.png 43 | 44 | .. raw:: html 45 | 46 |
47 | -------------------------------------------------------------------------------- /tutorial/en/beginner_source/examples_autograd/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/beginner_source/examples_autograd/README.txt -------------------------------------------------------------------------------- /tutorial/en/beginner_source/examples_nn/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/beginner_source/examples_nn/README.txt -------------------------------------------------------------------------------- /tutorial/en/beginner_source/examples_tensor/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/en/beginner_source/examples_tensor/README.txt -------------------------------------------------------------------------------- /tutorial/en/beginner_source/examples_tensor/two_layer_net_numpy.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | """ 3 | Warm-up: numpy 4 | -------------- 5 | 6 | A fully-connected ReLU network with one hidden layer and no biases, trained to 7 | predict y from x using Euclidean error. 8 | 9 | This implementation uses numpy to manually compute the forward pass, loss, and 10 | backward pass. 11 | 12 | A numpy array is a generic n-dimensional array; it does not know anything about 13 | deep learning or gradients or computational graphs, and is just a way to perform 14 | generic numeric computations. 15 | """ 16 | import numpy as np 17 | 18 | # N is batch size; D_in is input dimension; 19 | # H is hidden dimension; D_out is output dimension. 20 | N, D_in, H, D_out = 64, 1000, 100, 10 21 | 22 | # Create random input and output data 23 | x = np.random.randn(N, D_in) 24 | y = np.random.randn(N, D_out) 25 | 26 | # Randomly initialize weights 27 | w1 = np.random.randn(D_in, H) 28 | w2 = np.random.randn(H, D_out) 29 | 30 | learning_rate = 1e-6 31 | for t in range(500): 32 | # Forward pass: compute predicted y 33 | h = x.dot(w1) 34 | h_relu = np.maximum(h, 0) 35 | y_pred = h_relu.dot(w2) 36 | 37 | # Compute and print loss 38 | loss = np.square(y_pred - y).sum() 39 | print(t, loss) 40 | 41 | # Backprop to compute gradients of w1 and w2 with respect to loss 42 | grad_y_pred = 2.0 * (y_pred - y) 43 | grad_w2 = h_relu.T.dot(grad_y_pred) 44 | grad_h_relu = grad_y_pred.dot(w2.T) 45 | grad_h = grad_h_relu.copy() 46 | grad_h[h < 0] = 0 47 | grad_w1 = x.T.dot(grad_h) 48 | 49 | # Update weights 50 | w1 -= learning_rate * grad_w1 51 | w2 -= learning_rate * grad_w2 52 | -------------------------------------------------------------------------------- /tutorial/en/beginner_source/former_torchies/README.txt: -------------------------------------------------------------------------------- 1 | PyTorch for former Torch users 2 | ------------------------------ 3 | 4 | 1. tensor_tutorial.py 5 | Tensors 6 | http://pytorch.org/tutorials/beginner/former_torchies/tensor_tutorial.html 7 | 8 | 2. autograd.py 9 | Autograd 10 | http://pytorch.org/tutorials/beginner/former_torchies/autograd_tutorial.html 11 | 12 | 3. nn_tutorial.py 13 | nn package 14 | http://pytorch.org/tutorials/beginner/former_torchies/nn_tutorial.html 15 | 16 | 4. parallelism_tutorial.py 17 | Multi-GPU examples 18 | http://pytorch.org/tutorials/beginner/former_torchies/parallelism_tutorial.html 19 | -------------------------------------------------------------------------------- /tutorial/en/beginner_source/former_torchies_tutorial.rst: -------------------------------------------------------------------------------- 1 | PyTorch for former Torch users 2 | ------------------------------ 3 | **Author**: `Soumith Chintala `_ 4 | 5 | In this tutorial, you will learn the following: 6 | 7 | 1. Using torch Tensors, and important difference against (Lua)Torch 8 | 2. Using the autograd package 9 | 3. Building neural networks 10 | 11 | - Building a ConvNet 12 | - Building a Recurrent Net 13 | 14 | 4. Use multiple GPUs 15 | 16 | 17 | .. toctree:: 18 | :hidden: 19 | 20 | /beginner/former_torchies/tensor_tutorial 21 | /beginner/former_torchies/autograd_tutorial 22 | /beginner/former_torchies/nn_tutorial 23 | /beginner/former_torchies/parallelism_tutorial 24 | 25 | .. galleryitem:: /beginner/former_torchies/tensor_tutorial.py 26 | :figure: /_static/img/tensor_illustration_flat.png 27 | 28 | .. galleryitem:: /beginner/former_torchies/autograd_tutorial.py 29 | :figure: /_static/img/Variable.png 30 | 31 | .. galleryitem:: /beginner/former_torchies/nn_tutorial.py 32 | :figure: /_static/img/torch-nn-vs-pytorch-nn.png 33 | 34 | .. galleryitem:: /beginner/former_torchies/parallelism_tutorial.py 35 | 36 | .. raw:: html 37 | 38 |
-------------------------------------------------------------------------------- /tutorial/en/beginner_source/nlp/README.txt: -------------------------------------------------------------------------------- 1 | Deep Learning for NLP with Pytorch 2 | ---------------------------------- 3 | 4 | 1. pytorch_tutorial.py 5 | Introduction to PyTorch 6 | http://pytorch.org/tutorials/beginner/nlp/pytorch_tutorial.html 7 | 8 | 2. deep_learning_tutorial.py 9 | Deep Learning with PyTorch 10 | http://pytorch.org/tutorials/beginner/nlp/deep_learning_tutorial.html 11 | 12 | 3. word_embeddings_tutorial.py 13 | Word Embeddings: Encoding Lexical Semantics 14 | http://pytorch.org/tutorials/beginner/nlp/word_embeddings_tutorial.html 15 | 16 | 4. sequence_models_tutorial.py 17 | Sequence Models and Long-Short Term Memory Networks 18 | http://pytorch.org/tutorials/beginner/nlp/sequence_models_tutorial.html 19 | 20 | 5. advanced_tutorial.py 21 | Advanced: Making Dynamic Decisions and the Bi-LSTM CRF 22 | http://pytorch.org/tutorials/beginner/nlp/advanced_tutorial.html -------------------------------------------------------------------------------- /tutorial/en/build.sh: -------------------------------------------------------------------------------- 1 | # TODO: make sure pytorch installed 2 | pip install -r requirements.txt 3 | make docs 4 | -------------------------------------------------------------------------------- /tutorial/en/intermediate_source/README.txt: -------------------------------------------------------------------------------- 1 | Intermediate tutorials 2 | ---------------------- 3 | 4 | 1. char_rnn_classification_tutorial.py 5 | Classifying Names with a Character-Level RNN 6 | http://pytorch.org/tutorials/intermediate/char_rnn_classification_tutorial.html 7 | 8 | 2. char_rnn_generation_tutorial.py 9 | Generating Names with a Character-Level RNN 10 | http://pytorch.org/tutorials/intermediate/char_rnn_generation_tutorial.html 11 | 12 | 3. seq2seq_translation_tutorial.py 13 | Translation with a Sequence to Sequence Network and Attention 14 | http://pytorch.org/tutorials/intermediate/seq2seq_translation_tutorial.html 15 | 16 | 4. reinforcement_q_learning.py 17 | Reinforcement Learning (DQN) tutorial 18 | http://pytorch.org/tutorials/intermediate/reinforcement_q_learning.html 19 | 20 | 5. dist_tuto.rst 21 | Writing Distributed Applications with PyTorch 22 | http://pytorch.org/tutorials/intermediate/dist_tuto.html 23 | 24 | 6. spatial_transformer_tutorial 25 | Spatial Transformer Networks Tutorial 26 | http://pytorch.org/tutorials/intermediate/spatial_transformer_tutorial.html 27 | -------------------------------------------------------------------------------- /tutorial/en/requirements.txt: -------------------------------------------------------------------------------- 1 | sphinx 2 | sphinx_rtd_theme 3 | sphinx-gallery 4 | numpy 5 | matplotlib 6 | torchvision 7 | torch 8 | 9 | # to run examples 10 | pandas 11 | scikit-image 12 | pillow 13 | -------------------------------------------------------------------------------- /tutorial/zh/LICENSE: -------------------------------------------------------------------------------- 1 | BSD 3-Clause License 2 | 3 | Copyright (c) 2017, Pytorch contributors 4 | All rights reserved. 5 | 6 | Redistribution and use in source and binary forms, with or without 7 | modification, are permitted provided that the following conditions are met: 8 | 9 | * Redistributions of source code must retain the above copyright notice, this 10 | list of conditions and the following disclaimer. 11 | 12 | * Redistributions in binary form must reproduce the above copyright notice, 13 | this list of conditions and the following disclaimer in the documentation 14 | and/or other materials provided with the distribution. 15 | 16 | * Neither the name of the copyright holder nor the names of its 17 | contributors may be used to endorse or promote products derived from 18 | this software without specific prior written permission. 19 | 20 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 21 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 23 | DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE 24 | FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 26 | SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 27 | CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 28 | OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 29 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 | -------------------------------------------------------------------------------- /tutorial/zh/_static/img/SRResNet.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/SRResNet.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/Variable.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/Variable.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/cartpole.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/cartpole.gif -------------------------------------------------------------------------------- /tutorial/zh/_static/img/cat.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/cat.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/cat_224x224.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/cat_224x224.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/cat_output1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/cat_output1.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/char_rnn_generation.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/char_rnn_generation.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/cifar10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/cifar10.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/data_parallel.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/data_parallel.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/DistPyTorch.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/DistPyTorch.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/all_gather.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/all_gather.pdf -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/all_gather.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/all_gather.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/all_reduce.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/all_reduce.pdf -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/all_reduce.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/all_reduce.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/broadcast.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/broadcast.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/gather.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/gather.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/reduce.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/reduce.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/scatter.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/scatter.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/send_recv.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/send_recv.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/distributed/send_recv_big.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/distributed/send_recv_big.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/dynamic_graph.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/dynamic_graph.gif -------------------------------------------------------------------------------- /tutorial/zh/_static/img/landmarked_face2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/landmarked_face2.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/mnist.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/mnist.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/neural-style/dancing.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/neural-style/dancing.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/neural-style/neuralstyle.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/neural-style/neuralstyle.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/neural-style/picasso.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/neural-style/picasso.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/pytorch-logo-dark.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/pytorch-logo-dark.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/attention-decoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden;encoder_outputs} 11 | 12 | 13 | input -> embedding; 14 | embedding -> dropout; 15 | dropout -> embedded; 16 | 17 | embedded -> attn; 18 | prev_hidden -> attn; 19 | attn -> attn_softmax; 20 | attn_softmax -> attn_weights; 21 | attn_weights -> bmm; 22 | encoder_outputs -> bmm; 23 | bmm -> attn_applied; 24 | attn_applied -> attn_combine; 25 | embedded -> attn_combine; 26 | 27 | attn_combine -> relu -> gru; 28 | prev_hidden -> gru; 29 | gru -> out; 30 | gru -> hidden; 31 | 32 | out -> softmax; 33 | softmax -> output; 34 | 35 | {rank=same;output;hidden} 36 | 37 | // Layer nodes 38 | embedding [fillcolor=dodgerblue, fontcolor=white]; 39 | attn [fillcolor=dodgerblue, fontcolor=white]; 40 | attn_combine [fillcolor=dodgerblue, fontcolor=white]; 41 | bmm [fillcolor=dodgerblue, fontcolor=white]; 42 | gru [fillcolor=dodgerblue, fontcolor=white]; 43 | out [fillcolor=dodgerblue, fontcolor=white]; 44 | 45 | // Function nodes 46 | dropout [fillcolor=palegreen]; 47 | relu [fillcolor=palegreen]; 48 | softmax [fillcolor=palegreen]; 49 | attn_softmax [fillcolor=palegreen]; 50 | 51 | } 52 | -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/attention-decoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/attention-decoder-network.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/decoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden} 11 | 12 | input -> embedding; 13 | embedding -> relu; 14 | relu -> gru; 15 | 16 | prev_hidden -> gru; 17 | gru -> out; 18 | gru -> hidden; 19 | 20 | out -> softmax; 21 | softmax -> output; 22 | 23 | {rank=same;output;hidden} 24 | 25 | // Layer nodes 26 | embedding [fillcolor=dodgerblue, fontcolor=white]; 27 | gru [fillcolor=dodgerblue, fontcolor=white]; 28 | out [fillcolor=dodgerblue, fontcolor=white]; 29 | 30 | // Function nodes 31 | relu [fillcolor=palegreen]; 32 | softmax [fillcolor=palegreen]; 33 | 34 | } 35 | -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/decoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/decoder-network.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/decoder.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/decoder.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/decoder@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/decoder@2x.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/encoder-network.dot: -------------------------------------------------------------------------------- 1 | digraph G { 2 | 3 | // Main styles 4 | nodesep=0.3; ranksep=0.15; 5 | 6 | node [shape=rect, fillcolor=darkorange, color=white, style=filled, fontsize=11, fontname="arial", height=0.2]; 7 | edge [color=gray, arrowsize=0.5]; 8 | 9 | // Layout 10 | {rank=same;input;prev_hidden} 11 | 12 | input -> embedding; 13 | embedding -> embedded; 14 | embedded -> gru; 15 | prev_hidden -> gru; 16 | gru -> output; 17 | gru -> hidden; 18 | 19 | embedding [fillcolor=dodgerblue, fontcolor=white]; 20 | gru [fillcolor=dodgerblue, fontcolor=white]; 21 | 22 | } 23 | -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/encoder-network.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/encoder-network.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/seq2seq.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/seq2seq.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/seq2seq@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/seq2seq@2x.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/word-encoding.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/word-encoding.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq-seq-images/word-encoding@2x.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq-seq-images/word-encoding@2x.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/seq2seq_flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/seq2seq_flat.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/stn/FSeq.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/stn/FSeq.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/stn/Five.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/stn/Five.gif -------------------------------------------------------------------------------- /tutorial/zh/_static/img/stn/stn-arch.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/stn/stn-arch.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/stn/tr.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/stn/tr.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/tensor_illustration.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/tensor_illustration.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/tensor_illustration_flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/tensor_illustration_flat.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/thumbnails/babel.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/thumbnails/babel.jpg -------------------------------------------------------------------------------- /tutorial/zh/_static/img/thumbnails/default.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/thumbnails/default.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/thumbnails/examples.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/thumbnails/examples.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/thumbnails/pytorch-logo-flat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/thumbnails/pytorch-logo-flat.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/thumbnails/torch-logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/thumbnails/torch-logo.png -------------------------------------------------------------------------------- /tutorial/zh/_static/img/torch-nn-vs-pytorch-nn.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/_static/img/torch-nn-vs-pytorch-nn.png -------------------------------------------------------------------------------- /tutorial/zh/_templates/layout.html: -------------------------------------------------------------------------------- 1 | {% extends "!layout.html" %} 2 | 3 | {% block footer %} 4 | {{ super() }} 5 | 6 | 13 | 22 | 36 | {% endblock %} -------------------------------------------------------------------------------- /tutorial/zh/advanced_source/README.txt: -------------------------------------------------------------------------------- 1 | Advanced Tutorials 2 | ------------------ 3 | 4 | 1. neural_style_tutorial.py 5 | Neural Transfer with PyTorch 6 | http://pytorch.org/tutorials/advanced/neural_style_tutorial.html 7 | 8 | 2. numpy_extensions_tutorial.py 9 | Creating extensions using numpy and scipy 10 | http://pytorch.org/tutorials/advanced/numpy_extensions_tutorial.html 11 | 12 | 3. c_extension.rst 13 | Custom C extensions for pytorch 14 | http://pytorch.org/tutorials/advanced/c_extension.html 15 | 16 | 4. super_resolution_with_caffe2.py 17 | Transfering a model from PyTorch to Caffe2 and Mobile using ONNX 18 | http://pytorch.org/tutorials/advanced/super_resolution_with_caffe2.html 19 | -------------------------------------------------------------------------------- /tutorial/zh/apachecn-learning-group.rst: -------------------------------------------------------------------------------- 1 | 组织学习交流群 2 | ================== 3 | 4 | 机器学习交流群: `629470233 `__ （2000人） 5 | 6 | 大数据交流群: `214293307 `__ （2000人） 7 | 8 | Kaggle 竞赛交流群: `686932392 `__ （2000人） 9 | 10 | 了解我们: ``__ 11 | 12 | 加入组织: ``__ 13 | 14 | 更多学（zhuang）习（bi）交流群请参阅: ``__ -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/README.txt: -------------------------------------------------------------------------------- 1 | Beginner Tutorials 2 | ------------------ 3 | 4 | 1. blitz/* and deep_learning_60min_blitz.rst 5 | Deep Learning with PyTorch: A 60 Minute Blitz 6 | http://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html 7 | 8 | 2. former_torches/* and former_torchies_tutorial.rst 9 | PyTorch for former Torch users 10 | http://pytorch.org/tutorials/beginner/former_torchies_tutorial.html 11 | 12 | 3. examples_*/* and pytorch_with_examples.rst 13 | Learning PyTorch with Examples 14 | http://pytorch.org/tutorials/beginner/pytorch_with_examples.html 15 | 16 | 4. transfer_learning_tutorial.py 17 | Transfer Learning tutorial 18 | http://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html 19 | 20 | 5. nlp/* and deep_learning_nlp_tutorial.rst 21 | Deep Learning for NLP with Pytorch 22 | http://pytorch.org/tutorials/beginner/deep_learning_nlp_tutorial.html -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/blitz/README.txt: -------------------------------------------------------------------------------- 1 | Deep Learning with PyTorch: A 60 Minute Blitz 2 | --------------------------------------------- 3 | 4 | 1. tensor_tutorial.py 5 | What is PyTorch? 6 | http://pytorch.org/tutorials/beginner/blitz/tensor_tutorial.html 7 | 8 | 2. autograd_tutorial.py 9 | Autograd: automatic differentiation 10 | http://pytorch.org/tutorials/beginner/blitz/autograd_tutorial.html 11 | 12 | 3. neural_networks_tutorial.py 13 | Neural Networks 14 | http://pytorch.org/tutorials/beginner/blitz/neural_networks_tutorial.html# 15 | 16 | 4. cifar10_tutorial.py 17 | Training a classifier 18 | http://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/deep_learning_60min_blitz.rst: -------------------------------------------------------------------------------- 1 | PyTorch 深度学习: 60 分钟极速入门教程 2 | --------------------------------------------- 3 | **Author**: `Soumith Chintala `_ 4 | 5 | 本教程的目标: 6 | 7 | - 更高层次地理解 PyTorch 的 Tensor（张量）库以及神经网络. 8 | - 学会训练一个小的神经网络用来对图像进行分类 9 | 10 | *本教程假设您对 numpy 有基本的了解* 11 | 12 | .. Note:: 13 | 请确认您已经安装了 `torch`_ 和 `torchvision`_ 包. 14 | 15 | .. _torch: https://github.com/pytorch/pytorch 16 | .. _torchvision: https://github.com/pytorch/vision 17 | 18 | 19 | .. toctree:: 20 | :hidden: 21 | 22 | /beginner/blitz/tensor_tutorial 23 | /beginner/blitz/autograd_tutorial 24 | /beginner/blitz/neural_networks_tutorial 25 | /beginner/blitz/cifar10_tutorial 26 | /beginner/blitz/data_parallel_tutorial 27 | 28 | .. galleryitem:: /beginner/blitz/tensor_tutorial.py 29 | :figure: /_static/img/tensor_illustration_flat.png 30 | 31 | .. galleryitem:: /beginner/blitz/autograd_tutorial.py 32 | :figure: /_static/img/Variable.png 33 | 34 | .. galleryitem:: /beginner/blitz/neural_networks_tutorial.py 35 | :figure: /_static/img/mnist.png 36 | 37 | .. galleryitem:: /beginner/blitz/cifar10_tutorial.py 38 | :figure: /_static/img/cifar10.png 39 | 40 | .. galleryitem:: /beginner/blitz/data_parallel_tutorial.py 41 | :figure: /_static/img/data_parallel.png 42 | 43 | .. raw:: html 44 | 45 |
46 | -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/examples_autograd/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/beginner_source/examples_autograd/README.txt -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/examples_nn/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/beginner_source/examples_nn/README.txt -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/examples_tensor/README.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/wizardforcel/pytorch-doc-zh/a722cff892c7cc30fab952e6a7562742197c0f74/tutorial/zh/beginner_source/examples_tensor/README.txt -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/former_torchies/README.txt: -------------------------------------------------------------------------------- 1 | PyTorch for former Torch users 2 | ------------------------------ 3 | 4 | 1. tensor_tutorial.py 5 | Tensors 6 | http://pytorch.org/tutorials/beginner/former_torchies/tensor_tutorial.html 7 | 8 | 2. autograd.py 9 | Autograd 10 | http://pytorch.org/tutorials/beginner/former_torchies/autograd_tutorial.html 11 | 12 | 3. nn_tutorial.py 13 | nn package 14 | http://pytorch.org/tutorials/beginner/former_torchies/nn_tutorial.html 15 | 16 | 4. parallelism_tutorial.py 17 | Multi-GPU examples 18 | http://pytorch.org/tutorials/beginner/former_torchies/parallelism_tutorial.html 19 | -------------------------------------------------------------------------------- /tutorial/zh/beginner_source/former_torchies_tutorial.rst: -------------------------------------------------------------------------------- 1 | PyTorch for former Torch users 2 | ------------------------------ 3 | **Author**: `Soumith Chintala `_ 4 | 5 | 在本教程中，你将学习到以下内容: 6 | 7 | 1. 使用 torch Tensors, 它和 (Lua)Torch 有很大的不同 8 | 2. 使用 autograd package 9 | 3. 构建神经网络 10 | 11 | - 构建一个 ConvNet 12 | - 构建一个 Recurrent Net 13 | 14 | 4. 使用多个 GPUs 15 | 16 | 17 | .. toctree:: 18 | :hidden: 19 | 20 | /beginner/former_torchies/tensor_tutorial 21 | /beginner/former_torchies/autograd_tutorial 22 | /beginner/former_torchies/nn_tutorial 23 | /beginner/former_torchies/parallelism_tutorial 24 | 25 | .. galleryitem:: /beginner/former_torchies/tensor_tutorial.py 26 | :figure: /_static/img/tensor_illustration_flat.png 27 | 28 | .. galleryitem:: /beginner/former_torchies/autograd_tutorial.py 29 | :figure: /_static/img/Variable.png 30 | 31 | .. galleryitem:: /beginner/former_torchies/nn_tutorial.py 32 | :figure: /_static/img/torch-nn-vs-pytorch-nn.png 33 | 34 | .. galleryitem:: /beginner/former_torchies/parallelism_tutorial.py 35 | 36 | .. raw:: html 37 | 38 |
-------------------------------------------------------------------------------- /tutorial/zh/beginner_source/nlp/README.txt: -------------------------------------------------------------------------------- 1 | Deep Learning for NLP with Pytorch 2 | ---------------------------------- 3 | 4 | 1. pytorch_tutorial.py 5 | Introduction to PyTorch 6 | http://pytorch.org/tutorials/beginner/nlp/pytorch_tutorial.html 7 | 8 | 2. deep_learning_tutorial.py 9 | Deep Learning with PyTorch 10 | http://pytorch.org/tutorials/beginner/nlp/deep_learning_tutorial.html 11 | 12 | 3. word_embeddings_tutorial.py 13 | Word Embeddings: Encoding Lexical Semantics 14 | http://pytorch.org/tutorials/beginner/nlp/word_embeddings_tutorial.html 15 | 16 | 4. sequence_models_tutorial.py 17 | Sequence Models and Long-Short Term Memory Networks 18 | http://pytorch.org/tutorials/beginner/nlp/sequence_models_tutorial.html 19 | 20 | 5. advanced_tutorial.py 21 | Advanced: Making Dynamic Decisions and the Bi-LSTM CRF 22 | http://pytorch.org/tutorials/beginner/nlp/advanced_tutorial.html -------------------------------------------------------------------------------- /tutorial/zh/build.sh: -------------------------------------------------------------------------------- 1 | # TODO: make sure pytorch installed 2 | pip install -r requirements.txt 3 | make docs 4 | -------------------------------------------------------------------------------- /tutorial/zh/intermediate_source/README.txt: -------------------------------------------------------------------------------- 1 | Intermediate tutorials 2 | ---------------------- 3 | 4 | 1. char_rnn_classification_tutorial.py 5 | Classifying Names with a Character-Level RNN 6 | http://pytorch.org/tutorials/intermediate/char_rnn_classification_tutorial.html 7 | 8 | 2. char_rnn_generation_tutorial.py 9 | Generating Names with a Character-Level RNN 10 | http://pytorch.org/tutorials/intermediate/char_rnn_generation_tutorial.html 11 | 12 | 3. seq2seq_translation_tutorial.py 13 | Translation with a Sequence to Sequence Network and Attention 14 | http://pytorch.org/tutorials/intermediate/seq2seq_translation_tutorial.html 15 | 16 | 4. reinforcement_q_learning.py 17 | Reinforcement Learning (DQN) tutorial 18 | http://pytorch.org/tutorials/intermediate/reinforcement_q_learning.html 19 | 20 | 5. dist_tuto.rst 21 | Writing Distributed Applications with PyTorch 22 | http://pytorch.org/tutorials/intermediate/dist_tuto.html 23 | 24 | 6. spatial_transformer_tutorial 25 | Spatial Transformer Networks Tutorial 26 | http://pytorch.org/tutorials/intermediate/spatial_transformer_tutorial.html 27 | -------------------------------------------------------------------------------- /tutorial/zh/project-contributors.rst: -------------------------------------------------------------------------------- 1 | 项目贡献者 2 | ========== 3 | 4 | 该页面是 `apachecn/pytorch-doc-zh `__ 项目相关的贡献者信息. 5 | 6 | 项目负责人 7 | ---------- 8 | - `@那伊抹微笑 `__ 9 | 10 | 贡献者 11 | ------ 12 | 13 | PyTorch 中文教程贡献者 14 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 15 | 16 | - `@那伊抹微笑 `__ 17 | 18 | 建议反馈 19 | -------- 20 | 21 | - 联系项目负责人 `@那伊抹微笑 `__. 22 | - 在我们的 `apachecn/pytorch-doc-zh `__ github 上提 issue. 23 | - 发送邮件到 Email: pytorch#apachecn.org（#替换成@）. 24 | - 在我们的 `组织学习交流群 <./apachecn-learning-group.rst>`__ 中联系群主/管理员即可. 25 | -------------------------------------------------------------------------------- /tutorial/zh/requirements.txt: -------------------------------------------------------------------------------- 1 | sphinx 2 | sphinx_rtd_theme 3 | sphinx-gallery 4 | numpy 5 | matplotlib 6 | torchvision 7 | torch 8 | 9 | # to run examples 10 | pandas 11 | scikit-image 12 | pillow 13 | --------------------------------------------------------------------------------