Python API リファレンス
- 共通
- NdArray
NdArrayNdArray.bool_fill()NdArray.cast()NdArray.clear()NdArray.clear_calledNdArray.copy_from()NdArray.dataNdArray.data_ptr()NdArray.dtypeNdArray.fill()NdArray.from_numpy_array()NdArray.get_data()NdArray.masked_fill()NdArray.modification_countNdArray.narrow()NdArray.ndimNdArray.shapeNdArray.sizeNdArray.size_from_axis()NdArray.stridesNdArray.view()NdArray.zero()NdArray.zeroing
- Variable
VariableVariable.apply()Variable.backward()Variable.bool_fill_()Variable.clear_all_graph_links()Variable.dVariable.dataVariable.forward()Variable.from_numpy_array()Variable.function_referencesVariable.gVariable.get_number_of_referencesVariable.get_unlinked_variable()Variable.gradVariable.infoVariable.masked_fill_()Variable.ndimVariable.need_gradVariable.no_grad()Variable.parentVariable.persistentVariable.recomputeVariable.reset_shape()Variable.reshape()Variable.rewire_on()Variable.shapeVariable.sizeVariable.size_from_axis()Variable.unlinked()Variable.visit()Variable.visit_check()
- Computation Graph
- 関数
- Function
FunctionFunction.argsFunction.auto_grad_depends_input_data()Function.auto_grad_depends_output_data()Function.backward()Function.forward()Function.grad_depends_input_data()Function.grad_depends_output_data()Function.infoFunction.inplace_data()Function.inplace_data_with()Function.min_outputs()Function.need_setup_recompute()Function.recompute()Function.set_active_input_mask()Function.setup()Function.setup_recompute()Function.tags
PythonFunction
- 関数のリスト
- ニューラルネットワーク層
affine()convolution()depthwise_convolution()deconvolution()depthwise_deconvolution()deformable_convolution()adaptive_separable_convolution()max_pooling()average_pooling()global_average_pooling()sum_pooling()unpooling()embed()rnn()lstm()gru()multi_head_attention()patch_correlation()roi_align()
- ニューラルネットワークの活性化
- Normalization
batch_normalization()fused_batch_normalization()sync_batch_normalization()mean_subtraction()norm_normalization()clip_by_value()clip_grad_by_value()clip_by_norm()clip_grad_by_norm()layer_normalization()instance_normalization()group_normalization()weight_standardization()weight_normalization()spectral_norm()
- 削減
- 計算
- 論理
equal()equal_scalar()greater()greater_equal()greater_equal_scalar()greater_scalar()less()less_equal()less_equal_scalar()less_scalar()logical_and()logical_and_scalar()logical_not()logical_or()logical_or_scalar()logical_xor()logical_xor_scalar()not_equal()not_equal_scalar()sign()minimum2()maximum2()minimum_scalar()maximum_scalar()isnan()isinf()reset_nan()reset_inf()where()
- 数学
- 配列操作
concatenate()split()stack()slice()gather()gather_nd()scatter_nd()scatter_add()pad()transpose()broadcast()broadcast_to()tile()meshgrid()flip()shift()sort()shape()reshape()one_hot()assign()top_k_data()top_k_grad()pack_padded_sequence()pad_packed_sequence()searchsorted()bool_gather()bool_scatter()bool_fill()dot()
- 確率論
- Loss 関数
- 信号処理
- Geometric Neural Network Layers
- 量子化ニューラルネットワーク層
- サポートされていない、特別な使用
- 画像オブジェクト検出
- 検証
- ニューラルネットワーク層
- Function
- パラメトリック関数
- パラメータを管理するためのAPI
- パラメトリック関数リスト
parametric_function_api()affine()convolution()depthwise_convolution()deconvolution()depthwise_deconvolution()deformable_convolution()batch_normalization()fused_batch_normalization()sync_batch_normalization()mean_subtraction()layer_normalization()instance_normalization()group_normalization()rnn()lstm()gru()embed()prelu()svd_affine()svd_convolution()cpd3_convolution()binary_connect_affine()binary_connect_convolution()binary_weight_affine()binary_weight_convolution()inq_affine()inq_convolution()fixed_point_quantized_affine()fixed_point_quantized_convolution()min_max_quantized_affine()min_max_quantized_convolution()pow2_quantized_affine()pow2_quantized_convolution()pruned_affine()pruned_convolution()min_max_quantize()lstm_cell()LSTMCellspectral_norm()weight_normalization()multi_head_attention()transformer()transformer_encode()transformer_decode()
- パラメータイニシャライザー
- Grad
- Solvers
- Solver
SolverSolver.check_inf_grad()Solver.check_inf_or_nan_grad()Solver.check_nan_grad()Solver.clear_parameters()Solver.clip_grad_by_norm()Solver.get_parameters()Solver.get_states()Solver.infoSolver.learning_rate()Solver.load_states()Solver.nameSolver.remove_parameters()Solver.save_states()Solver.scale_grad()Solver.set_learning_rate()Solver.set_parameters()Solver.set_states()Solver.set_states_from_protobuf()Solver.set_states_to_protobuf()Solver.setup()Solver.update()Solver.weight_decay()Solver.weight_decay_is_fused()Solver.zero_grad()
- Solver のリスト
- Solver
- Callbacks
- Communicator
- Communicator のインターフェイス
CommunicatorCommunicator.abort()Communicator.add_context_and_parameters()Communicator.all_gather()Communicator.all_reduce()Communicator.all_reduce_callback()Communicator.allreduce()Communicator.barrier()Communicator.bcast()Communicator.clear_context_parameters()Communicator.find_group()Communicator.init()Communicator.list_groups()Communicator.local_rankCommunicator.nameCommunicator.new_group()Communicator.rankCommunicator.reduce()Communicator.reduce_scatter()Communicator.size
- Communicator の一覧
- Communicator のインターフェイス
- Monitor
- ユーティリティ
- NNP の保存やロードのユーティリティ
- 画像ユーティリティ
- データのイテレーター
- Debug Utils
- DLPack
- RNN Utils
- その他
- Quantization Aware Training
- QATConfig
QATConfigQATConfig.RecorderPositionQATConfig.RoundingMethodQATConfig.bn_foldingQATConfig.bn_self_foldingQATConfig.channel_lastQATConfig.channel_wiseQATConfig.dtypeQATConfig.ext_nameQATConfig.learning_rate_scaleQATConfig.narrow_rangeQATConfig.niter_to_recordingQATConfig.niter_to_trainingQATConfig.pow2QATConfig.record_layersQATConfig.recorder_activationQATConfig.recorder_positionQATConfig.recorder_weightQATConfig.round_modeQATConfig.skip_biasQATConfig.skip_inputs_layersQATConfig.skip_outputs_layersQATConfig.zero_point
- QATTensorRTConfig
- QATScheduler
- QATConfig
- Extension
- 学習済みモデル
- Out-of-core execution
- Modules
- Graph Definition
- Sequential
- Experimental
- Viewers
- Show Graph by Tensorboard
- Graph Converters
GraphConverterFunctionModifier- Function Modifiers
BatchNormalizationFoldingModifierAddBiasModifierBatchNormalizationFoldingModifierInnerBatchNormalizationFoldingOppositeModifierInnerBatchNormalizationSelfFoldingModifierFusedBatchNormalizationModifierUnfusedBatchNormalizationModifierChannelLastModifierChannelFirstModifierRemoveFunctionModifierBatchNormBatchStatModifierTestModeModifierIdentityModifierNoGradModifierPruningModifierQuantizeNonQNNToRecordingModifierQuantizeRecordingToTrainingModifier
- Trainers
- Mixed Precision Trainings
- DynamicLossScalingUpdater
DynamicLossScalingUpdaterDynamicLossScalingUpdater.solverDynamicLossScalingUpdater.lossDynamicLossScalingUpdater.data_feederDynamicLossScalingUpdater.scaleDynamicLossScalingUpdater.scaling_factorDynamicLossScalingUpdater.NDynamicLossScalingUpdater.clear_bufferDynamicLossScalingUpdater.accum_gradDynamicLossScalingUpdater.weight_decayDynamicLossScalingUpdater.commDynamicLossScalingUpdater.gradsDynamicLossScalingUpdater.update()
- DynamicLossScalingUpdater
- Parametric Function Classes