• Search by Module
• Search by Words
• Search Projects

• Java
• Python
• JavaScript
• TypeScript
• C++
• Scala
• Blog

• dragon-master
  • .github
    • ISSUE_TEMPLATE
      • bug-report.md
      • documentation-issue.md
  • cmake
    • FindNumPy.cmake
    • MiscCheck.cmake
    • Dependencies.cmake
    • FindProtobuf.cmake
    • SelectCudaArch.cmake
    • FindPythonLibs.cmake
    • StripDebugInfo.cmake
    • LinkLibrary.cmake
  • dragon
    • version.txt
    • proto
      • onnx.proto
      • CMakeLists.txt
      • dragon.proto
      • caffemodel.proto
    • utils
      • proto_utils.cc
      • map_utils.h
      • cuda_device.h
      • string_utils.h
      • dlpack.h
      • proto_utils.h
      • filler.h
      • omp_utils.h
      • macros.h
      • op_kernels.h
      • cudnn_device.h
      • thread.h
      • logging.cc
      • eigen_utils.h
      • caffemodel.h
      • cudnn_device.cc
      • CMakeLists.txt
      • math
        • random.cu
        • elementwise.h
        • reduce.cc
        • blas.h
        • elementwise.cu
        • blas.cu
        • reduce.cu
        • random.cc
        • random.h
        • blas.cc
        • utils.h
        • utils.cc
        • CMakeLists.txt
        • reduce.h
        • broadcast.cu
        • elementwise.cc
        • broadcast.cc
        • broadcast.h
      • logging.h
      • math_functions.h
      • cub_device.h
      • cast.h
    • kernels
      • recurrent
        • lstm_cell_op_kernel.cu
        • CMakeLists.txt
        • lstm_cell_op_kernel.cc
      • training
        • sgd_update_op_kernel.cc
        • adam_update_op_kernel.cu
        • sgd_update_op_kernel.cu
        • mprec_update_op_kernel.cc
        • nesterov_update_op_kernel.cu
        • rmsprop_update_op_kernel.cu
        • adam_update_op_kernel.cc
        • CMakeLists.txt
        • rmsprop_update_op_kernel.cc
        • mprec_update_op_kernel.cu
        • nesterov_update_op_kernel.cc
      • array
        • arange_op_kernel.cc
        • index_select_op_kernel.cc
        • one_hot_op_kernel.cu
        • tile_op_kernel.cc
        • masked_select_op_kernel.cc
        • cum_sum_op_kernel.cu
        • cast_op_kernel.cu
        • eye_op_kernel.cu
        • split_op_kernel.cu
        • concat_op_kernel.cu
        • transpose_op_kernel.cu
        • eye_op_kernel.cc
        • channel_shuffle_op_kernel.cu
        • split_op_kernel.cc
        • index_select_op_kernel.cu
        • repeat_op_kernel.cc
        • reduce_sum_op_kernel.cc
        • argreduce_op_kernel.cu
        • channel_normalize_op_kernel.cu
        • masked_select_op_kernel.cu
        • tile_op_kernel.cu
        • non_zero_op_kernel.cu
        • channel_normalize_op_kernel.cc
        • argreduce_op_kernel.cc
        • one_hot_op_kernel.cc
        • cast_op_kernel.cc
        • repeat_op_kernel.cu
        • CMakeLists.txt
        • pad_op_kernel.cc
        • transpose_op_kernel.cc
        • non_zero_op_kernel.cc
        • channel_shuffle_op_kernel.cc
        • slice_op_kernel.cc
        • reduce_sum_op_kernel.cu
        • concat_op_kernel.cc
        • slice_op_kernel.cu
        • pad_op_kernel.cu
        • cum_sum_op_kernel.cc
        • arange_op_kernel.cu
      • framework
        • gradient_op_kernel.cu
        • gradient_op_kernel.cc
        • CMakeLists.txt
      • control_flow
        • assign_op_kernel.cu
        • assign_op_kernel.cc
        • CMakeLists.txt
      • CMakeLists.txt
      • math
        • affine_op_kernel.cu
        • clip_op_kernel.cu
        • clip_op_kernel.cc
        • elementwise_op_kernel.cc
        • eltmentwise_op_kernel.cu
        • moments_op_kernel.cc
        • CMakeLists.txt
        • moments_op_kernel.cu
        • affine_op_kernel.cc
      • vision
        • maxpool_op_kernel.cc
        • resize_nearest_op_kernel.cc
        • roi_pool_op_kernel.cu
        • depthwise_conv_op_kernel.cu
        • resize_nearest_op_kernel.cu
        • roi_align_op_kernel.cc
        • conv_op_kernel.cc
        • bias_add_op_kernel.cu
        • avgpool_op_kernel.cc
        • roi_align_op_kernel.cu
        • avgpool_op_kernel.cu
        • bias_add_op_kernel.cc
        • conv_op_kernel.cu
        • roi_pool_op_kernel.cc
        • maxpool_op_kernel.cu
        • depthwise_conv_op_kernel.cc
        • CMakeLists.txt
        • resize_linear_op_kernel.cu
        • resize_linear_op_kernel.cc
      • loss
        • sigmoid_focal_loss_op_kernel.cc
        • sparse_softmax_ce_loss_op_kernel.cu
        • nll_loss_op_kernel.cu
        • generic_loss_op_kernel.cc
        • softmax_ce_loss_op_kernel.cc
        • nll_loss_op_kernel.cc
        • sigmoid_focal_loss_op_kernel.cu
        • CMakeLists.txt
        • generic_loss_op_kernel.cu
        • sigmoid_ce_loss_op_kernel.cc
        • softmax_ce_loss_op_kernel.cu
        • smooth_l1_loss_op_kernel.cu
        • sparse_softmax_ce_loss_op_kernel.cc
        • smooth_l1_loss_op_kernel.cc
        • sigmoid_ce_loss_op_kernel.cu
      • activation
        • drop_block_op_kernel.cu
        • sigmoid_op_kernel.cc
        • elu_op_kernel.cu
        • relu_op_kernel.cu
        • prelu_op_kernel.cu
        • elu_op_kernel.cc
        • prelu_op_kernel.cc
        • relu_op_kernel.cc
        • tanh_op_kernel.cu
        • drop_path_op_kernel.cc
        • dropout_op_kernel.cu
        • softmax_op_kernel.cc
        • drop_path_op_kernel.cu
        • softmax_op_kernel.cu
        • dropout_op_kernel.cc
        • selu_op_kernel.cc
        • CMakeLists.txt
        • sigmoid_op_kernel.cu
        • selu_op_kernel.cu
        • drop_block_op_kernel.cc
        • tanh_op_kernel.cc
      • normalization
        • group_norm_op_kernel.cu
        • lp_norm_op_kernel.cc
        • CMakeLists.txt
        • lp_norm_op_kernel.cu
        • batch_norm_op_kernel.cu
        • group_norm_op_kernel.cc
        • batch_norm_op_kernel.cc
    • onnx
      • onnx_backend.cc
      • onnx_initializer.cc
      • onnx_backend.h
      • CMakeLists.txt
      • onnx_attibute.cc
      • onnx_importer.cc
    • operators
      • recurrent
        • recurrent_op_cudnn.cc
        • recurrent_op.cc
        • recurrent_op.h
        • rnn_param_op.h
        • lstm_cell_op.h
        • rnn_param_op.cc
        • CMakeLists.txt
        • lstm_cell_op.cc
        • recurrent_op_cudnn.h
      • training
        • sgd_update_op.h
        • sgd_update_op.cc
        • nesterov_update_op.h
        • rmsprop_update_op.h
        • rmsprop_update_op.cc
        • nesterov_update_op.cc
        • adam_update_op.h
        • CMakeLists.txt
        • update_op_base.cc
        • adam_update_op.cc
        • update_op_base.h
      • distributed
        • collective_op.h
        • collective_op.cc
        • collective_op_base.h
        • CMakeLists.txt
      • array
        • split_op.cc
        • select_ops.h
        • cast_op.cc
        • cum_sum_op.cc
        • expand_dims_op.cc
        • channel_normalize_op.cc
        • reduce_min_op.cc
        • reshape_op.cc
        • transpose_op.cc
        • reduce_ops.h
        • tile_op.cc
        • repeat_op.h
        • stack_op.h
        • stack_op.cc
        • shape_op.cc
        • index_select_op.cc
        • channel_shuffle_op.cc
        • channel_normalize_op.h
        • slice_op.cc
        • argreduce_op.cc
        • reduce_mean_op.cc
        • initialize_ops.h
        • transpose_op.h
        • where_op.cc
        • tile_op.h
        • split_op.h
        • flatten_op.cc
        • slice_op.h
        • reduce_max_op.cc
        • masked_select_op.cc
        • one_hot_op.h
        • cast_op.h
        • shape_op.h
        • cum_ops.h
        • pad_op.cc
        • non_zero_op.cc
        • reduce_sum_op.cc
        • multinomial_op.cc
        • concat_op.cc
        • multinomial_op.h
        • CMakeLists.txt
        • repeat_op.cc
        • channel_shuffle_op.h
        • non_zero_op.h
        • argreduce_op.h
        • concat_op.h
        • reshape_ops.h
        • initialize_ops.cc
        • one_hot_op.cc
        • expand_op.cc
        • squeeze_op.cc
        • expand_op.h
        • pad_op.h
        • arange_op.cc
        • where_op.h
      • framework
        • gradient_ops.cc
        • gradient_ops.h
        • CMakeLists.txt
      • control_flow
        • masked_assign_op.cc
        • assign_op.cc
        • assign_ops.h
        • CMakeLists.txt
        • copy_op.h
        • copy_op.cc
      • CMakeLists.txt
      • math
        • log_op.cc
        • exp_op.cc
        • square_op.cc
        • matmul_op.h
        • fully_connected_op.h
        • sqrt_op.cc
        • accumulate.cc
        • cos_op.cc
        • elementwise_ops.cc
        • minimum_op.cc
        • pow_op.cc
        • accumulate_op.h
        • mul_op.cc
        • clip_op.h
        • moments_op.cc
        • moments_op.h
        • affine_op.cc
        • clip_op.cc
        • sub_op.cc
        • neg_op.cc
        • fully_connected_op.cc
        • reciprocal_op.cc
        • elementwise_ops.h
        • rsqrt_op.cc
        • dot_op.cc
        • abs_op.cc
        • dot_op.h
        • CMakeLists.txt
        • affine_op.h
        • add_op.cc
        • div_op.cc
        • sign_op.cc
        • sin_op.cc
        • maximum_op.cc
        • matmul_op.cc
      • metric
        • accuracy_op.cc
        • CMakeLists.txt
        • accuracy_op.h
      • vision
        • pool2d_op_cudnn.cc
        • roi_align_op.h
        • conv_op_base.h
        • conv2d_op.cc
        • bias_add_op.cc
        • roi_align_op.cc
        • pool_op_base.h
        • conv2d_transpose_op_cudnn.cc
        • pool_op_base.cc
        • depthwise_conv2d_op_cudnn.cc
        • conv2d_op_cudnn.cc
        • conv2d_transpose_op.cc
        • depth_to_space_op.cc
        • depthwise_conv2d_op.cc
        • conv_transpose_op.h
        • roi_pool_op.h
        • resize_op.h
        • conv_op_base.cc
        • depthwise_conv_op.h
        • CMakeLists.txt
        • roi_pool_op.cc
        • resize_op.cc
        • bias_add_op_cudnn.cc
        • pool_op.h
        • conv_op_cache.h
        • depth_to_space_op.h
        • space_to_depth_op.cc
        • space_to_depth_op.h
        • pool2d_op.cc
        • conv_op.h
        • bias_add_op.h
      • loss
        • softmax_ce_loss_op.cc
        • l1_loss_ops.h
        • l1_loss_op.cc
        • smooth_l1_loss_op.cc
        • ctc_loss_op.h
        • ctc_loss_op_cudnn.cc
        • sigmoid_loss_ops.h
        • ctc_loss_op.cc
        • sparse_softmax_ce_loss_op.cc
        • l2_loss_op.cc
        • CMakeLists.txt
        • sigmoid_focal_loss_op.cc
        • nll_loss_op.h
        • sigmoid_ce_loss_op.cc
        • l2_loss_op.h
        • softmax_loss_ops.h
        • nll_loss_op.cc
      • activation
        • tanh_op_cudnn.cc
        • sigmoid_op.h
        • prelu_op.h
        • relu_op_cudnn.cc
        • dropout_op.cc
        • softmax_op.h
        • elu_op.h
        • drop_path_op.h
        • selu_op.h
        • dropout_op.h
        • sigmoid_op.cc
        • tanh_op.cc
        • drop_block2d_op.cc
        • softmax_op_cudnn.cc
        • relu_op.cc
        • sigmoid_op_cudnn.cc
        • tanh_op.h
        • CMakeLists.txt
        • drop_path_op.cc
        • relu_op.h
        • prelu_op.cc
        • elu_op_cudnn.cc
        • selu_op.cc
        • softmax_op.cc
        • elu_op.cc
        • dropout_op_cudnn.cc
        • drop_block_op.h
      • normalization
        • batch_norm_op.cc
        • lp_normalize_op.h
        • lrn_op_cudnn.cc
        • group_norm_op.h
        • lrn_op.h
        • lp_normalize_op.cc
        • CMakeLists.txt
        • group_norm_op.cc
        • lrn_op.cc
        • batch_norm_op_cudnn.cc
        • batch_norm_op_sync.cc
        • batch_norm_op.h
    • setup.py
    • python
      • _api
        • dlpack
          • __init__.py
        • distributed
          • __init__.py
        • cuda
          • __init__.py
        • nn
          • __init__.py
        • losses
          • __init__.py
        • updaters
          • __init__.py
        • io
          • __init__.py
        • workspace
          • __init__.py
        • __init__.py
        • math
          • __init__.py
        • random
          • __init__.py
        • autograph
          • __init__.py
        • vision
          • __init__.py
        • metrics
          • __init__.py
        • bitwise
          • __init__.py
        • logging
          • __init__.py
      • utils
        • __init__.py
        • vision
          • data_transformer.py
          • data_iterator.py
          • __init__.py
      • vm
        • onnx
          • frontend.py
          • shell.py
          • exporter.py
          • serialization.py
          • nodes
            • activation.py
            • vision.py
            • normalization.py
            • array.py
            • __init__.py
            • math.py
          • __init__.py
          • utils.py
          • helper.py
        • __init__.py
      • __init__.py
      • tools
        • tensorboard.py
        • __init__.py
        • cpp_extension.py
      • backend.py
      • core
        • proto
          • tf_example.proto
          • __init__.py
          • dragon.proto
        • eager
          • dlpack.py
          • backprop.py
          • executor.py
          • __init__.py
          • context.py
          • tensor.py
        • training
          • sgd.py
          • adam.py
          • rmsprop.py
          • __init__.py
          • updater.py
        • distributed
          • __init__.py
          • backend.py
        • util
          • decorator.py
          • six.py
          • math_util.py
          • inspect.py
          • tls.py
          • nest.py
          • registry.py
          • deprecation.py
          • __init__.py
          • logging.py
        • framework
          • workspace.py
          • config.py
          • ops.py
          • mapping.py
          • tensor_spec.py
          • __init__.py
          • context.py
          • proto_util.py
          • types.py
          • device_spec.py
        • io
          • kpl_record.py
          • tf_record.py
          • reader.py
          • __init__.py
        • __init__.py
        • ops
          • distributed_ops_lib.py
          • training_ops_lib.py
          • loss_ops_lib.py
          • tensorbind_symbolic.py
          • math_ops_lib.py
          • vision_ops_lib.py
          • distributed_ops.py
          • normalization_ops.py
          • init_ops.py
          • control_flow_ops_lib.py
          • activation_ops_lib.py
          • metric_ops.py
          • control_flow_ops.py
          • rnn_ops.py
          • tensorbind_eager.py
          • activation_ops.py
          • array_ops_lib.py
          • metric_ops_lib.py
          • vision_ops.py
          • __init__.py
          • math_ops.py
          • utils.py
          • framework_ops.py
          • rnn_ops_lib.py
          • loss_ops.py
          • array_ops.py
          • init_ops_lib.py
          • normalization_ops_lib.py
        • autograph
          • grad_impl.py
          • def_function.py
          • config.py
          • op_def.py
          • grad_maker.py
          • function_lib.py
          • op_spec.py
          • __init__.py
          • tensor.py
        • device
          • cuda.py
          • __init__.py
    • CMakeLists.txt
    • requirements.txt
    • core
      • graph.cc
      • common.h
      • workspace.cc
      • registry.h
      • graph_optimizer.cc
      • operator_schema.h
      • context.h
      • graph_gradient.cc
      • graph.h
      • context_cuda.h
      • graph_gradient.h
      • context_cnml.h
      • graph_optimizer.h
      • types.h
      • memory.cc
      • memory.h
      • CMakeLists.txt
      • context.cc
      • typeid.h
      • workspace.h
      • tensor.h
      • operator_gradient.h
      • operator_schema.cc
      • operator.cc
      • operator.h
    • modules
      • runtime
        • module.cc
        • dragon_runtime.h
        • CMakeLists.txt
        • device.cc
      • python
        • config.h
        • common.h
        • dlpack.h
        • proto.h
        • module.cc
        • plugin_op.cc
        • plugin_op.h
        • types.h
        • cuda.h
        • autograd.h
        • mpi.h
        • CMakeLists.txt
        • tensor.h
        • operator.h
  • dali
    • _api
      • __init__.py
      • ops
        • __init__.py
    • plugin
      • pytorch.py
      • __init__.py
    • __init__.py
    • core
      • iterator.py
      • pipeline.py
      • __init__.py
      • ops
        • geometric.py
        • decoder.py
        • random.py
        • paste.py
        • builtin.py
        • fused.py
        • color.py
        • reader.py
        • resize.py
        • crop.py
        • array.py
        • __init__.py
      • context.py
      • types.py
  • third_party
    • mpi
      • README.md
      • build.sh
    • pybind11
    • protobuf
      • README.md
    • cub
    • eigen
  • .gitmodules
  • LICENSE
  • .clang-format
  • .flake8
  • docker
    • ubuntu-16.04-cpu
      • Dockerfile
    • README.md
    • ubuntu-16.04-cuda9.0-cudnn7
      • Dockerfile
  • torch
    • utils
      • dlpack.py
      • hooks.py
      • model_zoo.py
      • __init__.py
    • optim
      • optimizer.py
      • sgd.py
      • adam.py
      • rmsprop.py
      • __init__.py
    • onnx
      • __init__.py
      • utils.py
    • distributed
      • __init__.py
    • autograd
      • function.py
      • variable.py
      • __init__.py
      • grad_mode.py
    • jit
      • tape.py
      • trace.py
      • __init__.py
    • executor.py
    • nn
      • _reduction.py
      • init.py
      • functional.py
      • __init__.py
      • parameter.py
      • modules
        • module.py
        • container.py
        • batchnorm.py
        • activation.py
        • conv.py
        • padding.py
        • normalization.py
        • loss.py
        • _functions.py
        • dropout.py
        • pooling.py
        • affine.py
        • __init__.py
        • linear.py
        • utils.py
        • rnn.py
        • upsampling.py
    • cpp.py
    • serialization.py
    • __init__.py
    • ops
      • training
        • functional.py
        • _functions.py
        • __init__.py
      • init
        • functional.py
        • _functions.py
        • __init__.py
      • distributed
        • functional.py
        • _functions.py
        • __init__.py
      • array
        • functional.py
        • _functions.py
        • __init__.py
      • tensorbind.py
      • __init__.py
      • math
        • functional.py
        • _functions.py
        • __init__.py
      • utils.py
      • metric
        • functional.py
        • _functions.py
        • __init__.py
    • vision
      • models
        • vgg.py
        • alexnet.py
        • squeezenet.py
        • resnet.py
        • __init__.py
        • inception.py
      • __init__.py
      • ops
        • functional.py
        • _functions.py
        • __init__.py
    • tensor.py
  • README.md
  • scripts
    • build_windows_vc14.bat
    • build_windows_vc12.bat
    • build_windows_vc142.bat
    • build_arm64-v8a.sh
    • build_windows_vc141.bat
  • tools
    • codegen_runtime.py
  • tensorlayer
    • _api
      • layers
        • __init__.py
      • initializers
        • __init__.py
      • activations
        • __init__.py
      • models
        • __init__.py
      • losses
        • __init__.py
      • __init__.py
    • __init__.py
    • core
      • layers
        • activation.py
        • convolution
          • simplified_conv.py
          • conv_utils.py
          • __init__.py
        • normalization.py
        • inputs.py
        • shape.py
        • pooling.py
        • dense
          • base_dense.py
          • __init__.py
        • __init__.py
        • utils.py
        • core.py
        • merge.py
      • files
        • __init__.py
        • utils.py
      • activations.py
      • models
        • resnet.py
        • __init__.py
      • losses.py
      • engine
        • module.py
        • container.py
        • layer.py
        • model.py
        • node.py
        • __init__.py
      • __init__.py
      • initializers.py
  • tensorflow
    • _api
      • nn
        • __init__.py
      • __init__.py
      • math
        • __init__.py
      • random
        • __init__.py
      • keras
        • layers
          • __init__.py
        • initializers
          • __init__.py
        • activations
          • __init__.py
        • losses
          • __init__.py
        • regularizers
          • __init__.py
        • __init__.py
        • optimizers
          • __init__.py
      • linalg
        • __init__.py
      • bitwise
        • __init__.py
      • dtypes
        • __init__.py
    • __init__.py
    • core
      • proto
        • types.proto
        • config.proto
        • __init__.py
      • eager
        • def_function.py
        • backprop.py
        • __init__.py
      • training
        • __init__.py
        • tracking
          • layer_utils.py
          • __init__.py
      • module
        • module.py
        • __init__.py
      • framework
        • constant_op.py
        • ops.py
        • tensor_spec.py
        • __init__.py
        • framework_lib.py
        • tensor_shape.py
        • dtypes.py
      • __init__.py
      • ops
        • standard_ops.py
        • nn.py
        • linalg_ops.py
        • init_ops.py
        • variables.py
        • random_ops.py
        • losses
          • loss_reduction.py
          • __init__.py
        • gradients_impl.py
        • __init__.py
        • math_ops.py
        • nn_impl.py
        • nn_ops.py
        • array_ops.py
        • bitwise_ops.py
        • clip_ops.py
      • keras
        • regularizers.py
        • utils
          • generic_utils.py
          • conv_utils.py
          • __init__.py
          • losses_utils.py
        • layers
          • convolutional.py
          • normalization.py
          • pooling.py
          • __init__.py
          • core.py
          • advanced_activations.py
          • merge.py
        • saving
          • pickle_format.py
          • __init__.py
        • activations.py
        • losses.py
        • engine
          • base_layer.py
          • sequential.py
          • training.py
          • input_spec.py
          • __init__.py
          • network.py
          • input_layer.py
        • optimizer
          • optimizer.py
          • gradient_descent.py
          • adam.py
          • rmsprop.py
          • __init__.py
        • __init__.py
        • initializers.py
  • .gitignore
  • docs
    • api
      • python
        • dragon
          • cuda.rst
          • reset_workspace.rst
          • nn.rst
          • dlpack
            • from_dlpack.rst
            • to_dlpack.rst
          • masked_select.rst
          • stack.rst
          • logging.rst
          • cast.rst
          • device.rst
          • name_scope.rst
          • autograph.rst
          • io.rst
          • expand_dims.rst
          • metrics.rst
          • vision.rst
          • nonzero.rst
          • math.rst
          • channel_normalize.rst
          • distributed
            • broadcast.rst
            • get_world_size.rst
            • get_backend.rst
            • init.rst
            • is_initialized.rst
            • new_group.rst
            • all_reduce.rst
            • get_group.rst
            • get_rank.rst
          • losses.rst
          • shape.rst
          • workspace.rst
          • stop_gradient.rst
          • one_hot.rst
          • cuda
            • synchronize.rst
            • get_device_capability.rst
            • enable_cudnn.rst
            • set_default_device.rst
            • enable_cudnn_benchmark.rst
            • current_device.rst
            • is_available.rst
            • set_device.rst
          • graph_mode.rst
          • ones_like.rst
          • zeros.rst
          • Tensor.rst
          • index_select.rst
          • tile.rst
          • fill.rst
          • assign.rst
          • ones.rst
          • nn
            • layer_norm.rst
            • GRU.rst
            • elu.rst
            • prelu.rst
            • depthwise_conv2d.rst
            • instance_norm.rst
            • conv2d.rst
            • LSTM.rst
            • log_softmax.rst
            • dropout.rst
            • drop_block2d.rst
            • RNN.rst
            • bias_add.rst
            • drop_path.rst
            • space_to_depth.rst
            • local_response_norm.rst
            • fully_connected.rst
            • pool2d.rst
            • depth_to_space.rst
            • group_norm.rst
            • batch_norm.rst
            • relu.rst
            • softmax.rst
            • conv2d_transpose.rst
            • relu6.rst
            • sync_batch_norm.rst
            • selu.rst
            • leaky_relu.rst
          • eye_like.rst
          • updaters.rst
          • pad.rst
          • eye.rst
          • losses
            • ctc_loss.rst
            • sigmoid_focal_loss.rst
            • softmax_cross_entropy.rst
            • sparse_softmax_cross_entropy.rst
            • l2_loss.rst
            • l1_loss.rst
            • sigmoid_cross_entropy.rst
            • nll_loss.rst
            • smooth_l1_loss.rst
          • concat.rst
          • random.rst
          • slice.rst
          • updaters
            • Adam.rst
            • SGD.rst
            • Nesterov.rst
            • RMSProp.rst
          • python_plugin.rst
          • squeeze.rst
          • create_function.rst
          • Workspace_.rst
          • channel_shuffle.rst
          • copy.rst
          • dlpack.rst
          • arange.rst
          • eager_scope.rst
          • constant.rst
          • load_library.rst
          • eager_mode.rst
          • io
            • TFRecordWriter.rst
            • KPLRecordDataset.rst
            • KPLRecordWriter.rst
            • DataReader.rst
            • TFRecordExample.rst
          • workspace
            • save.rst
            • run_operator.rst
            • fetch_tensor.rst
            • feed_tensor.rst
            • has_tensor.rst
            • reset_tensor.rst
            • load.rst
          • gradients.rst
          • math
            • mul.rst
            • log.rst
            • mean.rst
            • sqrt.rst
            • matmul.rst
            • sub.rst
            • maximum.rst
            • equal.rst
            • tanh.rst
            • not_equal.rst
            • floor.rst
            • reciprocal.rst
            • is_nan.rst
            • greater_equal.rst
            • moving_average.rst
            • sin.rst
            • max.rst
            • cumsum.rst
            • rsqrt.rst
            • minimum.rst
            • clip.rst
            • square.rst
            • greater.rst
            • exp.rst
            • pow.rst
            • moments.rst
            • is_inf.rst
            • ceil.rst
            • less_equal.rst
            • abs.rst
            • lp_normalize.rst
            • min.rst
            • sum.rst
            • add.rst
            • negative.rst
            • less.rst
            • affine.rst
            • argmin.rst
            • div.rst
            • accumulate.rst
            • cos.rst
            • round.rst
            • dot.rst
            • sigmoid.rst
            • sign.rst
            • argmax.rst
          • TensorSpec.rst
          • split.rst
          • where.rst
          • zeros_like.rst
          • transpose.rst
          • random
            • uniform.rst
            • glorot_uniform.rst
            • set_seed.rst
            • normal_like.rst
            • truncated_normal.rst
            • glorot_normal.rst
            • uniform_like.rst
            • multinomial.rst
            • normal.rst
          • masked_assign.rst
          • autograph
            • set_execution.rst
            • set_optimization.rst
            • set_scheduler.rst
            • set_verbosity.rst
          • reshape.rst
          • bitwise.rst
          • vision
            • roi_align.rst
            • roi_pool.rst
            • DataIterator.rst
            • resize.rst
          • distributed.rst
          • metrics
            • accuracy.rst
          • repeat.rst
          • function.rst
          • get_workspace.rst
          • EagerTensor.rst
          • bitwise
            • bitwise_or.rst
            • invert.rst
            • bitwise_and.rst
            • bitwise_xor.rst
          • flatten.rst
          • logging
            • info.rst
            • debug.rst
            • warning.rst
            • fatal.rst
            • set_directory.rst
            • error.rst
            • set_verbosity.rst
          • GradientTape.rst
          • broadcast_to.rst
        • Makefile
        • make.bat
        • dragon.rst
        • tensorrt.rst
        • dali
          • device.rst
          • ops.rst
          • get_device_type.rst
          • get_distributed_info.rst
          • Pipeline.rst
          • Iterator.rst
          • ops
            • CropMirrorNormalize.rst
            • BBoxPaste.rst
            • Slice.rst
            • ExternalSource.rst
            • BbFlip.rst
            • ImageDecoder.rst
            • ImageDecoderRandomCrop.rst
            • Reshape.rst
            • Pad.rst
            • KPLRecordReader.rst
            • Cast.rst
            • RandomBBoxCrop.rst
            • Uniform.rst
            • Resize.rst
            • Paste.rst
            • TFRecordReader.rst
            • CoinFlip.rst
            • BrightnessContrast.rst
            • Hsv.rst
        • onnx
          • Shell.rst
        • onnx.rst
        • tensorflow.rst
        • torch
          • ne.rst
          • mul.rst
          • log.rst
          • mean.rst
          • no_grad.rst
          • nn.rst
          • uniform.rst
          • sqrt.rst
          • masked_select.rst
          • stack.rst
          • set_grad_enabled.rst
          • sub.rst
          • Size.rst
          • utils
            • dlpack
              • from_dlpack.rst
              • to_dlpack.rst
            • dlpack.rst
          • maximum.rst
          • device.rst
          • optim
            • Optimizer.rst
            • Adam.rst
            • SGD.rst
            • RMSprop.rst
          • expand.rst
          • cat.rst
          • floor.rst
          • reciprocal.rst
          • chunk.rst
          • onnx
            • export.rst
          • nonzero.rst
          • channel_normalize.rst
          • distributed
            • broadcast.rst
            • all_reduce.rst
          • jit
            • trace.rst
          • sin.rst
          • one_hot.rst
          • max.rst
          • rand.rst
          • ones_like.rst
          • zeros.rst
          • cumsum.rst
          • bitwise_xor.rst
          • index_select.rst
          • onnx.rst
          • gt.rst
          • ones.rst
          • nn
            • GRU.rst
            • ReplicationPad2d.rst
            • ReplicationPad3d.rst
            • functional
              • max_pool2d.rst
              • mse_loss.rst
              • ctc_loss.rst
              • avg_pool2d.rst
              • elu.rst
              • cross_entropy.rst
              • prelu.rst
              • conv_transpose2d.rst
              • sigmoid_focal_loss.rst
              • depthwise_conv2d.rst
              • tanh.rst
              • binary_cross_entropy_with_logits.rst
              • interpolate.rst
              • conv2d.rst
              • normalize.rst
              • log_softmax.rst
              • l1_loss.rst
              • pad.rst
              • dropout.rst
              • drop_block2d.rst
              • drop_path.rst
              • local_response_norm.rst
              • linear.rst
              • upsample.rst
              • group_norm.rst
              • batch_norm.rst
              • affine.rst
              • nll_loss.rst
              • relu.rst
              • softmax.rst
              • smooth_l1_loss.rst
              • softmax_cross_entropy_with_logits.rst
              • upsample_nearest.rst
              • relu6.rst
              • sync_batch_norm.rst
              • upsample_bilinear.rst
              • selu.rst
              • sigmoid.rst
              • leaky_relu.rst
            • L1Loss.rst
            • NLLLoss.rst
            • UpsamplingNearest2d.rst
            • Upsample.rst
            • ZeroPad2d.rst
            • GumbelSoftmax.rst
            • CrossEntropyLoss.rst
            • DropBlock2d.rst
            • SyncBatchNorm.rst
            • SCEWithLogitsLoss.rst
            • LogSoftmax.rst
            • LSTM.rst
            • Affine.rst
            • AvgPool2d.rst
            • BatchNorm1d.rst
            • Module.rst
            • Dropout.rst
            • LocalResponseNorm.rst
            • Tanh.rst
            • ReplicationPad1d.rst
            • MaxPool2d.rst
            • RNN.rst
            • ReflectionPad1d.rst
            • Sigmoid.rst
            • ReflectionPad2d.rst
            • SELU.rst
            • ELU.rst
            • UpsamplingBilinear2d.rst
            • BCEWithLogitsLoss.rst
            • BatchNorm3d.rst
            • CTCLoss.rst
            • DepthwiseConv2d.rst
            • ReflectionPad3d.rst
            • DropPath.rst
            • LSTMCell.rst
            • ReLU6.rst
            • ConstantPad1d.rst
            • Linear.rst
            • Softmax.rst
            • PReLU.rst
            • SigmoidFocalLoss.rst
            • Parameter.rst
            • BatchNorm2d.rst
            • ConvTranspose2d.rst
            • GroupNorm.rst
            • functional.rst
            • Conv2d.rst
            • ConstantPad3d.rst
            • ReLU.rst
            • LeakyReLU.rst
            • MSELoss.rst
            • SmoothL1Loss.rst
            • ConstantPad2d.rst
          • isnan.rst
          • bitwise_not.rst
          • rsqrt.rst
          • minimum.rst
          • tensor.rst
          • eye.rst
          • clamp.rst
          • exp.rst
          • squeeze.rst
          • topk.rst
          • eq.rst
          • lt.rst
          • pow.rst
          • optim.rst
          • ceil.rst
          • multinomial.rst
          • channel_shuffle.rst
          • empty.rst
          • abs.rst
          • arange.rst
          • min.rst
          • permute.rst
          • sum.rst
          • mm.rst
          • add.rst
          • le.rst
          • enable_grad.rst
          • from_numpy.rst
          • logsumexp.rst
          • split.rst
          • where.rst
          • zeros_like.rst
          • argmin.rst
          • normal.rst
          • jit.rst
          • reshape.rst
          • div.rst
          • vision
            • ops.rst
            • ops
              • roi_align.rst
              • roi_pool.rst
          • accumulate.rst
          • Tensor_.rst
          • isinf.rst
          • distributed.rst
          • cos.rst
          • repeat.rst
          • round.rst
          • narrow.rst
          • topk_acc.rst
          • randn.rst
          • ge.rst
          • sign.rst
          • unsqueeze.rst
          • argmax.rst
        • index.rst
        • conf.py
        • tensorlayer
          • layers
            • Flatten.rst
            • Relu.rst
            • MeanPool2d.rst
            • GlobalMaxPool2d.rst
            • Layer.rst
            • Transpose.rst
            • GlobalMeanPool2d.rst
            • BatchNorm.rst
            • Reshape.rst
            • LayerList.rst
            • MaxPool2d.rst
            • Concat.rst
            • Input.rst
            • Dense.rst
            • Conv2d.rst
            • Elementwise.rst
          • initializers
            • Constant.rst
            • TruncatedNormal.rst
            • Ones.rst
            • GlorotUniform.rst
            • RandomNormal.rst
            • GlorotNormal.rst
            • RandomUniform.rst
            • Zeros.rst
            • Initializer.rst
          • layers.rst
          • initializers.rst
          • models
            • Model.rst
          • models.rst
        • dali.rst
        • tensorflow
          • linalg.rst
          • optimizers.rst
          • nn.rst
          • clip_by_value.rst
          • cast.rst
          • device.rst
          • name_scope.rst
          • expand_dims.rst
          • gather.rst
          • math.rst
          • losses.rst
          • shape.rst
          • range.rst
          • one_hot.rst
          • initializers.rst
          • ones_like.rst
          • zeros.rst
          • fill.rst
          • identity.rst
          • ones.rst
          • nn
            • max_pool2d.rst
            • avg_pool2d.rst
            • elu.rst
            • depthwise_conv2d.rst
            • batch_normalization.rst
            • conv2d.rst
            • max_pool.rst
            • log_softmax.rst
            • dropout.rst
            • conv_transpose.rst
            • moments.rst
            • local_response_normalization.rst
            • space_to_depth.rst
            • depth_to_space.rst
            • relu.rst
            • convolution.rst
            • softmax.rst
            • conv2d_transpose.rst
            • relu6.rst
            • avg_pool.rst
            • selu.rst
            • leaky_relu.rst
          • pad.rst
          • eye.rst
          • concat.rst
          • random.rst
          • slice.rst
          • squeeze.rst
          • constant.rst
          • gradients.rst
          • math
            • log.rst
            • sqrt.rst
            • l2_normalize.rst
            • equal.rst
            • tanh.rst
            • add_n.rst
            • not_equal.rst
            • floor.rst
            • reciprocal.rst
            • is_nan.rst
            • greater_equal.rst
            • sin.rst
            • divide.rst
            • cumsum.rst
            • rsqrt.rst
            • square.rst
            • greater.rst
            • exp.rst
            • pow.rst
            • is_inf.rst
            • ceil.rst
            • less_equal.rst
            • abs.rst
            • multiply.rst
            • add.rst
            • negative.rst
            • less.rst
            • argmin.rst
            • subtract.rst
            • reduce_min.rst
            • cos.rst
            • reduce_max.rst
            • round.rst
            • reduce_mean.rst
            • reduce_sum.rst
            • sigmoid.rst
            • sign.rst
            • argmax.rst
          • TensorSpec.rst
          • split.rst
          • dtypes.rst
          • zeros_like.rst
          • transpose.rst
          • random
            • uniform.rst
            • truncated_normal.rst
            • normal.rst
          • reshape.rst
          • bitwise.rst
          • keras.rst
          • keras
            • optimizers.rst
            • activations.rst
            • layers
              • Flatten.rst
              • Layer.rst
              • DepthwiseConv2D.rst
              • GlobalAveragePooling2D.rst
              • Subtract.rst
              • GlobalMaxPool2D.rst
              • Maximum.rst
              • Add.rst
              • Reshape.rst
              • Conv2DTranspose.rst
              • Permute.rst
              • Dropout.rst
              • SELU.rst
              • ELU.rst
              • Concatenate.rst
              • AveragePooling2D.rst
              • Conv2D.rst
              • BatchNormalization.rst
              • Minimum.rst
              • Softmax.rst
              • Multiply.rst
              • Dense.rst
              • ReLU.rst
              • LeakyReLU.rst
              • MaxPool2D.rst
            • initializers
              • Constant.rst
              • TruncatedNormal.rst
              • Ones.rst
              • GlorotUniform.rst
              • RandomNormal.rst
              • GlorotNormal.rst
              • RandomUniform.rst
              • VarianceScaling.rst
              • Zeros.rst
              • Initializer.rst
            • activations
              • exponential.rst
              • elu.rst
              • tanh.rst
              • linear.rst
              • relu.rst
              • softmax.rst
              • selu.rst
              • sigmoid.rst
            • losses.rst
            • layers.rst
            • initializers.rst
            • losses
              • SparseCategoricalCrossentropy.rst
              • Loss.rst
              • sparse_categorical_crossentropy.rst
              • mean_absolute_error.rst
              • MeanAbsoluteError.rst
              • binary_crossentropy.rst
              • BinaryCrossentropy.rst
              • categorical_crossentropy.rst
              • MeanSquaredError.rst
              • CategoricalCrossentropy.rst
              • mean_squared_error.rst
            • Sequential.rst
            • regularizers
              • l2.rst
              • L1L2.rst
              • Regularizer.rst
              • l1_l2.rst
              • l1.rst
            • Input.rst
            • optimizers
              • Optimizer.rst
              • Adam.rst
              • SGD.rst
              • RMSprop.rst
            • regularizers.rst
          • linalg
            • matmul.rst
          • function.rst
          • bitwise
            • bitwise_or.rst
            • invert.rst
            • bitwise_and.rst
            • bitwise_xor.rst
          • dtypes
            • DType.rst
            • as_dtype.rst
          • GradientTape.rst
          • broadcast_to.rst
        • torch.rst
        • caffe.rst
        • caffe
          • AdamSolver.rst
          • layers
            • SoftmaxWithLoss.rst
            • Normalize.rst
            • Flatten.rst
            • Pooling.rst
            • Power.rst
            • EuclideanLoss.rst
            • BatchNorm.rst
            • FusedBatchNorm.rst
            • Deconvolution.rst
            • Reshape.rst
            • Permute.rst
            • LRN.rst
            • FusedGroupNorm.rst
            • Dropout.rst
            • Cast.rst
            • Python.rst
            • ROIPooling.rst
            • Sigmoid.rst
            • Convolution.rst
            • Eltwise.rst
            • Scale.rst
            • SELU.rst
            • ELU.rst
            • Concat.rst
            • Crop.rst
            • DepthwiseConv2d.rst
            • ROIAlign.rst
            • Input.rst
            • Data.rst
            • Tile.rst
            • Softmax.rst
            • PReLU.rst
            • SigmoidCrossEntropyLoss.rst
            • GroupNorm.rst
            • TanH.rst
            • ReLU.rst
            • SmoothL1Loss.rst
            • InnerProduct.rst
            • ArgMax.rst
            • StopGradient.rst
            • Accuracy.rst
            • Reduction.rst
          • SGDSolver.rst
          • layers.rst
          • RMSPropSolver.rst
          • NesterovSolver.rst
          • Net.rst
        • tensorrt
          • Binding.rst
          • backend
            • prepare.rst
            • supports_device.rst
            • run_node.rst
            • run_model.rst
          • Engine.rst
          • ONNXBackendRep.rst
          • backend.rst
      • _static
        • images
        • favicon.ico
      • cc
        • Doxyfile
        • images
        • template
          • layout.xml
          • footer.html
          • header.html
        • css
          • main.css
          • stylesheet.css
    • README.md
  • caffe
    • proto
      • __init__.py
      • caffe.proto
    • solver.py
    • layers
      • data.py
      • vision.py
      • loss.py
      • common.py
      • neuron.py
      • __init__.py
    • layer.py
    • net_spec.py
    • net.py
    • __init__.py
  • tensorrt
    • python
      • _api
        • backend
          • __init__.py
        • __init__.py
      • __init__.py
      • core
        • engine.py
        • __init__.py
        • backend.py

Dragon is a C(Computation)G(Graph)V(Virtual)M(Machine) based distributed deep learning framework. It fuses several modern frameworks and integrations together, powered by a unified engine. The computation between different programming styles is deterministic and reproduceable.

It is the first deep learning framework that focuses on developing multiple styles, rather than promoting internal interfaces. We will always learn from the AI community to evolve Dragon over time.

Installation

See the install guide for the pip package or how to build from source.

License

BSD 2-Clause license