Skip to main content
torch.js has not been released yet.
torch.js logotorch.js logotorch.js
PlaygroundContact
Login
Documentation
IntroductionType SafetyTensor ExpressionsTensor IndexingEinsumEinopsAutogradTraining a ModelProfiling & MemoryPyTorch MigrationBest PracticesRuntimesPerformancePyTorch CompatibilityBenchmarksDType Coverage
adaptive_avg_pool1dadaptive_avg_pool2dadaptive_avg_pool3dadaptive_max_pool1dadaptive_max_pool1d_with_indicesadaptive_max_pool2dadaptive_max_pool2d_with_indicesadaptive_max_pool3dadaptive_max_pool3d_with_indicesAdaptiveMaxPoolFunctionalOptionsaffine_gridAffineGridFunctionalOptionsalpha_dropoutAlphaDropoutFunctionalOptionsavg_pool1davg_pool2davg_pool3dAvgPool1dFunctionalOptionsAvgPool2dFunctionalOptionsAvgPool3dFunctionalOptionsbatch_normBatchNormFunctionalOptionsbinary_cross_entropybinary_cross_entropy_with_logitsBinaryCrossEntropyFunctionalOptionsBinaryCrossEntropyWithLogitsFunctionalOptionsCeluFunctionalOptionschannel_shuffleconv_transpose1dconv_transpose2dconv_transpose3dconv1dConv1dFunctionalOptionsconv2dConv2dFunctionalOptionsconv3dConv3dFunctionalOptionsConvTranspose1dFunctionalOptionsConvTranspose2dFunctionalOptionsConvTranspose3dFunctionalOptionscosine_embedding_losscosine_similarityCosineEmbeddingLossFunctionalOptionsCosineSimilarityFunctionalOptionscross_entropyCrossEntropyFunctionalOptionsctc_lossCTCLossOptionsdropoutdropout1ddropout2ddropout3dDropoutFunctionalOptionsEluFunctionalOptionsembeddingembedding_bagEmbeddingBagFunctionalOptionsEmbeddingFunctionalOptionsfeature_alpha_dropoutfoldFoldFunctionalOptionsfractional_max_pool2dfractional_max_pool2d_with_indicesfractional_max_pool3dfractional_max_pool3d_with_indicesFractionalMaxPoolFunctionalOptionsgaussian_nll_lossGluFunctionalOptionsgrid_sampleGridSampleFunctionalOptionsgroup_normgrouped_mmGroupedMMFunctionalOptionsGroupNormFunctionalOptionsHardshrinkFunctionalOptionsHardtanhFunctionalOptionshinge_embedding_lossHingeEmbeddingLossFunctionalOptionshuber_lossHuberLossFunctionalOptionsinstance_normInstanceNormFunctionalOptionsinterpolateInterpolateFunctionalOptionskl_divKlDivFunctionalOptionsKLDivOptionsl1_lossL1LossFunctionalOptionslayer_normLayerNormFunctionalOptionsLeakyReluFunctionalOptionslinearlocal_response_normLocalResponseNormFunctionalOptionslog_softmaxlp_pool1dlp_pool2dlp_pool3dLPPoolFunctionalOptionsmargin_ranking_lossMarginRankingLossFunctionalOptionsmax_pool1dmax_pool1d_with_indicesmax_pool2dmax_pool2d_with_indicesmax_pool3dmax_pool3d_with_indicesmax_unpool1dmax_unpool2dmax_unpool3dMaxPool1dFunctionalOptionsMaxPool2dFunctionalOptionsMaxPool3dFunctionalOptionsMaxUnpoolFunctionalOptionsmse_lossMseLossFunctionalOptionsmulti_head_attention_forwardmulti_margin_lossMultiHeadAttentionFunctionalOptionsmultilabel_margin_lossmultilabel_soft_margin_lossnll_lossNllLossFunctionalOptionsnormalizeNormalizeFunctionalOptionsone_hotpadPadFunctionalOptionspairwise_distancePairwiseDistanceFunctionalOptionspdistPdistFunctionalOptionspixel_shufflepixel_unshufflepoisson_nll_lossPoolWithIndicesResultReluFunctionalOptionsrms_normRmsNormFunctionalOptionsRreluFunctionalOptionsscaled_grouped_mmscaled_mmScaledDotProductAttentionFunctionalOptionsScaledGroupedMMFunctionalOptionsScaledMMFunctionalOptionssmooth_l1_lossSmoothL1LossFunctionalOptionssoft_margin_lossSoftMarginLossFunctionalOptionsSoftmaxOptionsSoftminFunctionalOptionsSoftplusFunctionalOptionsSoftshrinkFunctionalOptionstriplet_margin_losstriplet_margin_with_distance_lossTripletMarginLossFunctionalOptionsunfoldUnfoldFunctionalOptionsupsampleupsample_bilinearupsample_nearestUpsampleBilinearOptionsUpsampleNearestOptionsUpsampleOptions
ActivationOptionsAdaptiveAvgPool1dAdaptiveAvgPool2dAdaptiveAvgPool3dAdaptiveLogSoftmaxOptionsAdaptiveLogSoftmaxWithLossAdaptiveMaxPool1dAdaptiveMaxPool1dOptionsAdaptiveMaxPool2dAdaptiveMaxPool2dOptionsAdaptiveMaxPool3dAdaptiveMaxPool3dOptionsadd_moduleAlphaDropoutappendappendapplyAvgPool1dAvgPool1dOptionsAvgPool2dAvgPool2dOptionsAvgPool3dAvgPool3dOptionsBackwardHookBackwardPreHookBatchNorm1dBatchNorm2dBatchNorm3dBatchNormOptionsBCELossBCEWithLogitsLossBilinearBilinearOptionsBufferBufferOptionsBufferRegistrationHookbufferscallCELUCELUOptionsChannelShufflechildrenCircularPad1dCircularPad2dCircularPad3dclearConstantPad1dConstantPad2dConstantPad3dConv1dConv2dConv3dConvOptionsConvTranspose1dConvTranspose2dConvTranspose3dConvTransposeOptionsCosineEmbeddingLossCosineEmbeddingLossOptionsCosineSimilarityCosineSimilarityOptionscreatecreateCrossEntropyLossCTCLossdecodedecodedeleteDropoutDropout1dDropout2dDropout3dDropoutOptionsELUELUOptionsEmbeddingEmbeddingBagEmbeddingBagForwardOptionsEmbeddingBagFromPretrainedOptionsEmbeddingBagOptionsEmbeddingFromPretrainedOptionsEmbeddingOptionsencodeencodeentriesentriesevalextendFeatureAlphaDropoutFlattenFlattenOptionsFoldFoldOptionsforwardforwardforwardforwardforwardforwardforwardforwardforwardforwardforwardforwardforward_with_targetForwardHookForwardPreHookFractionalMaxPool2dFractionalMaxPool3dFractionalMaxPoolOptionsfrom_pretrainedfrom_pretrainedGaussianNLLLossGELUGELUOptionsgenerate_square_subsequent_maskgetgetgetgetgetget_bufferget_parameterget_submoduleGLUGLUOptionsGroupNormGroupNormOptionsGRUGRUCellHardshrinkHardshrinkOptionsHardsigmoidHardswishHardtanhHardtanhOptionshashasHingeEmbeddingLossHingeEmbeddingLossOptionsHuberLossHuberLossOptionsIdentityInstanceNorm1dInstanceNorm2dInstanceNorm3dInstanceNormOptionsis_uninitialized_bufferis_uninitialized_parameteriterator]iterator]iterator]iterator]keyskeysKLDivLossL1LossL1LossOptionsLayerNormLayerNormOptionsLazyBatchNorm1dLazyBatchNorm2dLazyBatchNorm3dLazyConv1dLazyConv2dLazyConv3dLazyConvOptionsLazyConvTranspose1dLazyConvTranspose2dLazyConvTranspose3dLazyConvTransposeOptionsLazyInstanceNorm1dLazyInstanceNorm2dLazyInstanceNorm3dLazyLinearLeakyReLULeakyReLUOptionsLinearLinearOptionsload_state_dictload_state_dictLocalResponseNormLocalResponseNormOptionslog_probLogSigmoidLogSoftmaxLogSoftmaxOptionsLPPool1dLPPool1dOptionsLPPool2dLPPool2dOptionsLPPool3dLPPool3dOptionsLSTMLSTMCellLSTMCellOptionsMarginRankingLossMarginRankingLossOptionsmaterializematerializematerialize_uninitializedmaterialize_uninitializedMaxPool1dMaxPool1dOptionsMaxPool2dMaxPool2dOptionsMaxPool3dMaxPool3dOptionsMaxUnpool1dMaxUnpool1dOptionsMaxUnpool2dMaxUnpool2dOptionsMaxUnpool3dMaxUnpool3dOptionsMishModuleModuleBuffersModuleChildrenModuleDictModuleDictOptionsModuleListModuleListOptionsModuleParametersModuleRegistrationHookmodulesMSELossMSELossOptionsmultihead_attnMultiheadAttentionMultiheadAttentionOptionsMultiheadAttnOptionsMultiLabelMarginLossMultiLabelMarginLossOptionsMultiLabelSoftMarginLossMultiMarginLossnamed_buffersnamed_childrennamed_modulesnamed_parametersNamedModulesOptionsNamedRecurseOptionsNLLLossnum_parametersNumParametersOptionsPairwiseDistancePairwiseDistanceOptionsParameterParameterDictParameterDictOptionsParameterListParameterListOptionsParameterOptionsParameterRegistrationHookparametersPixelShufflePixelUnshufflePoissonNLLLosspoppopPReLUPReLUOptionsRecurseOptionsReflectionPad1dReflectionPad2dReflectionPad3dregister_backward_hookregister_bufferregister_forward_hookregister_forward_pre_hookregister_full_backward_hookregister_full_backward_pre_hookregister_module_backward_hookregister_module_buffer_registration_hookregister_module_forward_hookregister_module_forward_pre_hookregister_module_full_backward_hookregister_module_full_backward_pre_hookregister_module_module_registration_hookregister_module_parameter_registration_hookregister_parameterReLUReLU6RemovableHandleremoveReplicationPad1dReplicationPad2dReplicationPad3dRMSNormRMSNormOptionsRNNRNNBaseRNNBaseOptionsRNNCellRNNCellOptionsRReLURReLUOptionsrunrunSELUSequentialsetsetsetSigmoidSiLUSmoothL1LossSmoothL1LossOptionsSoftMarginLossSoftMarginLossOptionsSoftmaxSoftmax2dSoftmaxOptionsSoftminSoftminOptionsSoftplusSoftplusOptionsSoftshrinkSoftshrinkOptionsSoftsignstate_dictstate_dictStateDictOptionsstepSyncBatchNormTanhTanhshrinkThresholdThresholdOptionstotototrainTrainOptionsTransformerTransformerDecoderTransformerDecoderDecodeOptionsTransformerDecoderLayerTransformerDecoderLayerDecodeOptionsTransformerDecoderLayerOptionsTransformerDecoderOptionsTransformerEncoderTransformerEncoderEncodeOptionsTransformerEncoderLayerTransformerEncoderLayerEncodeOptionsTransformerEncoderLayerOptionsTransformerEncoderOptionsTransformerOptionsTransformerRunOptionsTripletMarginLossTripletMarginWithDistanceLossUnflattenUnfoldUnfoldOptionsUninitializedBufferUninitializedOptionsUninitializedParameterupdateUpsampleUpsamplingBilinear2dUpsamplingNearest2dvaluesvalueszero_gradZeroPad1dZeroPad2dZeroPad3d
absacosacoshAdaptivePool1dShapeAdaptivePool2dShapeaddaddbmmAddbmmOptionsaddcdivAddcdivOptionsaddcmulAddcmulOptionsaddmmAddmmOptionsaddmvAddmvOptionsaddrAddrOptionsadjointallallcloseAllcloseOptionsAlphaBetaOptionsamaxaminaminmaxAminmaxOptionsangleanyapplyOutarangeare_deterministic_algorithms_enabledargmaxargminargsortargwhereas_stridedas_tensorasinasinhAssertNoShapeErrorAssertNotErrorAsStridedOptionsAtat_error_index_out_of_boundsatanatan2atanhatleast_1datleast_2datleast_3dAtShapeautocast_decrement_nestingautocast_increment_nestingautograd_gradient_mismatch_errorautograd_not_registered_errorAutogradConfigAutogradDeviceAutogradDTypeAutogradEntryAutogradHandleAutogradHandleImplAxesRecordBackwardFnbaddbmmBaddbmmOptionsbartlett_windowBaseKernelConfigbatch_dimensions_do_not_match_errorbernoulliBernoulliOptionsBinaryBackwardFnBinaryBroadcastResultBinaryDTypeBinaryKernelConfigCPUBinaryKernelCPUBinaryOpConfigBinaryOpNamesBinaryOpSchemaBinaryOptionsbincountBincountOptionsbitwise_andbitwise_left_shiftbitwise_notbitwise_orbitwise_right_shiftbitwise_xorblackman_windowblock_diagbmmBooleanDTypeRulebroadcast_error_incompatible_dimensionsbroadcast_shapesbroadcast_tensorsbroadcast_toBroadcastShapeBroadcastShapeRulebroadcastShapesbucketizeBucketizeOptionsBufferUsagebuildEinopsErrorbuildErrorMessagecanBroadcastTocartesian_prodcatCatOptionsCatShapeCauchyOptionscdistCdistOptionsceilceluCeluFunctionalOptionschain_matmulCheckShapeErrorCholeskyShapechunkchunk_error_dim_out_of_rangeChunkOptionsclampClampOptionsclear_autocast_cacheclearEinopsCacheclearEinsumCacheclonecolumn_stackcombinationsCombinationsOptionscompiled_with_cxx11_abicomplexconjconj_physicalcontiguousConv1dShapeConv2dShapeConv3dShapeConvTranspose2dShapecopysigncorrcoefcoscoshcount_nonzeroCountNonzeroOptionscovcoverage_reportcoverageReportCoverageReportCovOptionsCPUForwardFnCPUKernelConfigCPUKernelEntryCPUOnlyResultCPUTensorDatacreateCumExtremeResultcreateTorchCreationOpSchemaCumExtremeResultcummaxcummincumprodCumShapecumsumcumulative_trapezoidCumulativeOptionsCumulativeOptionsWithDimdeg2raddetachDeterministicOptionsDetShapeDevicedevice_error_requiresDeviceBufferDeviceCapabilitiesDeviceCheckedResultDeviceConfigDeviceContextDeviceEntryDeviceHandleDeviceInputDeviceOptionsDeviceRegistryDeviceTypediagdiag_embedDiagEmbedOptionsdiagflatDiagflatOptionsDiagFlatOptionsdiagonal_scatterDiagonalOptionsDiagonalScatterOptionsDiagOptionsDiagShapediffDiffOptionsdigammadimension_error_out_of_rangeDispatchConfigdistDistOptionsdivdotDotShapeRuleDoubleDoubleDimdropoutDropoutFunctionalOptionsdsplitdstackDTypedtype_already_registered_errordtype_components_mismatch_errordtype_not_found_errorDTypeComponentsDTypeConfigDTypeCoverageReportDTypeDisplayConfigDTypeEntryDTypeHandleDTypeHandleImplDTypeInfoDTypeRegistryDTypeRuleDTypeSerializationConfigDynamicShapeEigShapeeinops_error_ambiguous_decompositioneinops_error_anonymous_in_outputeinops_error_dimension_mismatcheinops_error_invalid_patterneinops_error_reduce_undefined_outputeinops_error_repeat_missing_sizeeinops_error_undefined_axiseinsumeinsum_error_dimension_mismatcheinsum_error_index_out_of_rangeeinsum_error_invalid_equationeinsum_error_invalid_sublist_elementeinsum_error_operand_count_mismatcheinsum_error_subscript_rank_mismatcheinsum_error_unknown_output_indexEinsumOptionsEinsumOutputShapeEllipsiseluelu_EluFunctionalOptionsembedding_bag_error_requires_2d_inputemptyempty_cacheempty_likeeqequalerferfcerfinvexpexp2expandexpand_asexpand_error_incompatibleExpandShapeexpm1ExponentialOptionseyeEyeOptionsfftFFTOptionsfindKernelWithPredicatefindSimilarPatternsflattenFlattenOptionsFlattenShapeflipflip_error_dim_out_of_rangefliplrFlipShapeflipudfloat_powerFloatDTypeRulefloorfloor_dividefmaxfminfmodformatEquationErrorformatShapefracfrexpfrombufferfullfull_likefunction_already_registered_errorFunctionConfigFunctionEntryFunctionHandlegathergather_error_dim_out_of_rangeGatherShapegcdgegeluGeometricOptionsget_autocast_cpu_dtypeget_autocast_gpu_dtypeget_autocast_ipu_dtypeget_autocast_xla_dtypeget_default_deviceget_default_dtypeget_deterministic_debug_modeget_device_configget_device_contextget_device_moduleget_dtype_infoget_file_pathget_float32_matmul_precisionget_num_interop_threadsget_num_threadsget_op_infoget_printoptionsget_real_dtypeget_rng_stategetAutogradgetDTypegetEinopsCacheSizegetEinsumCacheSizegetFunctiongetKernelgetMethodgetOpInfoGetOpKindGetOpSchemagetScalarKernelgluGluFunctionalOptionsGradContextGradFnGradientsForgtHalfHalfDimhamming_windowhann_windowhardshrinkhardsigmoidhardswishhardtanhhardtanh_HardtanhFunctionalOptionshas_autogradhas_devicehas_dtypehas_kernelhasAutogradhasDTypehasFunctionhasKernelhasMethodhasScalarKernelHasShapeErrorheavisidehistcHistcOptionshistogramHistogramOptionsHistogramResulthsplithstackhypoti0IdentityShapeifftimagindex_addindex_copyindex_fillindex_putindex_reduceindex_selectindex_select_error_dim_out_of_rangeIndexPutOptionsIndexSelectShapeIndexSpecIndicesOptionsIndicesSpecinitialize_deviceInputsForInsertDiminvalid_config_errorinverseInverseShapeirfftis_anomaly_check_nan_enabledis_anomaly_enabledis_autocast_cache_enabledis_autocast_cpu_enabledis_autocast_ipu_enabledis_autocast_xla_enabledis_complexis_complex_dtypeis_cpu_only_modeis_deterministic_algorithms_warn_only_enabledis_floating_pointis_floating_point_dtypeis_inference_mode_enabledis_nonzerois_tensoris_warn_always_enabledis_webgpu_availableIs2DIsAtLeast1DIsBinaryOpIsBinaryOpNameiscloseIscloseOptionsisfiniteisinisinfisnanisneginfisposinfisrealIsReductionOpIsReductionOpNameIsRegistryErrorIsShapeErroristftISTFTOptionsIsUnaryOpIsUnaryOpNameitem_error_not_scalarItemResultkaiser_windowKaiserWindowOptionskernel_not_registered_errorkernel_signature_mismatch_errorKernelConfigKernelConfigWebGPUKernelEntryKernelHandleKernelInfoKernelPredicateKernelRegistryKernelWebGPUkronkthvalueKthvalueOptionslcmldexpleleaky_reluleaky_relu_LeakyReluFunctionalOptionslerplevenshteinDistancelgammalinalg_error_not_square_matrixlinalg_error_requires_2dlinalg_error_requires_at_least_2dlinearlinspacelist_custom_deviceslist_custom_dtypeslist_deviceslist_dtypeslist_functionslist_kernelslist_methodslist_opslistCustomDTypeslistDTypeslistFunctionslistKernelsListKernelsOptionslistMethodslistOpsListOpsOptionsloglog_softmaxlog10log1plog2logaddexplogaddexp2logcumsumexplogical_andlogical_notlogical_orlogical_xorLogitOptionsLogNormalOptionsLogOptionslogsigmoidlogspacelogsumexpLogsumexpOptionsltLUShapeLuSolveOptionsmasked_fillmasked_selectmasked_select_asyncMaskSpecmatmulmatmul_error_inner_dimensions_do_not_matchMatmul2DShapeMatmulShapeMatmulShapeRuleMatrixTransposeShapemaxmaximummeanmedianmemory_statsmemory_summarymeshgridmethod_already_registered_errormethod_dtype_not_supported_errorMethodConfigMethodEntryMethodHandleminminimummishmmMMShapeRulemodemovedimmsortmulmultinomialmultinomial_asyncMultinomialAsyncOptionsMultinomialOptionsMultiplyBymvMVShapeRulenan_to_numnanmeannanmediannanquantileNanReductionOptionsnansumNanToNumOptionsnarrownarrow_copynarrow_error_length_exceeds_boundsnarrow_error_start_out_of_boundsNarrowShapeneneedsBroadcastnegNegativeDimnextafternonzeroNonzeroOptionsnormnormalNormalOptionsNormOptionsnumelonesones_likeop_kind_mismatch_errorop_not_found_errorOpCoverageEntryOpInfoOpKindOpNameOpSchemaOpSchemasouterOuterShapepackPackShapepermutepermute_error_dimension_count_mismatchPermuteShapepoissonpolarPool1dShapePool2dShapePool3dShapepositivepowpreluPrintOptionsprodprofiler_allow_cudagraph_cupti_lazy_reinit_cuda12promote_typesPromoteDTypeRulePutOptionsquantileQuantileOptionsrad2degrandrand_likerandintrandint_likeRandintLikeOptionsRandintOptionsrandnrandn_likeRandomLikeOptionsRandomOptionsrandpermRangeSpecRankravelrealrearrangeRearrangeOptionsRearrangeShapereciprocalreduceReduceOperationReduceOptionsReduceShapeReductionKernelConfigCPUReductionKernelCPUReductionOpNamesReductionOpSchemaReductionOptionsReductionShapeRuleregister_backwardregister_deviceregister_dtyperegister_forwardregister_functionregister_methodregister_scalar_forwardregisterAutogradRegisterBackwardOptionsregisterBinaryOpregisterDTypeRegisterDTypeOptionsRegisteredDTyperegisterFunctionRegisterFunctionOptionsregisterKernelRegisterKernelOptionsregisterMethodRegisterMethodOptionsregisterScalarKernelregisterUnaryOpregistration_failed_errorrelurelu_relu6ReluFunctionalOptionsremainderRemoveDimrepeatrepeat_interleaveRepeatInterleaveOptionsRepeatOptionsRepeatShapeReplaceDimrequireWebGPUreset_peak_memory_statsreshapeReshapeShaperesult_typerfftrollRollOptionsrot90Rot90Optionsroundrrelurrelu_RreluFunctionalOptionsrsqrtSafeExpandShapeSameDTypeRuleSameShapeRuleSaveForBackwardScalarCPUForwardFnScalarCPUKernelConfigScalarKernelEntryScalarKernelHandleScalarWebGPUKernelConfigScaleDimscatterscatter_addscatter_add_scatter_error_dim_out_of_rangescatter_reducescatter_reduce_ScatterReduceOptionsScatterShapesearchsortedSearchSortedOptionsselectselect_error_index_out_of_boundsselect_scatterSelectShapeseluset_default_deviceset_default_tensor_typeset_deterministic_debug_modeset_float32_matmul_precisionset_printoptionsset_warn_alwaysSetupContextFnShapeShapeCheckedResultShapedTensorShapeErrorMessageShapeOpSchemaShapeRulesigmoidsignsignbitsilusinsincsinhSizeOptionsslice_error_out_of_boundsslice_scatterSliceOptionsSliceScatterOptionsSliceShapeSliceSpecsoftmaxsoftmax_error_dim_out_of_rangeSoftmaxShapesoftminSoftminFunctionalOptionssoftplusSoftplusFunctionalOptionssoftshrinksoftsignsortSortOptionssplitsplit_error_dim_out_of_rangeSplitOptionssqrtsquaresqueezeSqueezeOptionsSqueezeShapestackStackOptionsStackShapestdstd_meanStdVarMeanOptionsStdVarOptionsstftSTFTOptionsStrideOptionssubSublistSublistElementSubscriptIndexsumSVDShapeswapaxessym_floatsym_intsym_notttaketake_along_dimTakeAlongDimOptionstantanhtanhshrinktensortensor_splitTensorCreatorTensorDatatensordotTensordotOptionsTensorLikeTensorMetaTensorOptionsTensorStoragethresholdthreshold_tileTileShapeToOptionstopkTopkOptionsTorchtraceTraceShapetransposetranspose_dims_error_out_of_rangetranspose_error_requires_2d_tensorTransposeDimsShapeTransposeDimsShapeCheckedTransposeShapetrapezoidTrapezoidOptionsTriangularOptionstriltril_indicesTriOptionsTripletriutriu_indicestrue_dividetruncTupleOfLengthTypedArrayTypedArrayForTypedStorageTypeOptionsUnaryBackwardFnUnaryDTypeUnaryKernelConfigCPUUnaryKernelCPUUnaryOpConfigUnaryOpFnUnaryOpNamesUnaryOpParamsUnaryOpSchemaUnaryOptionsunbindunbind_error_dim_out_of_rangeUnbindOptionsunflattenUniformOptionsuniqueunique_consecutiveUniqueConsecutiveOptionsUniqueOptionsunpackUnpackShapeunravel_indexunregister_deviceunsqueezeUnsqueezeOptionsUnsqueezeShapeuse_deterministic_algorithmsValidateBatchedSquareMatrixValidateChunkDimValidatedEinsumShapevalidateDeviceValidateDeviceValidatedRearrangeShapeValidatedReduceShapeValidatedRepeatShapevalidateDTypeValidateEinsumValidateOperandCountValidateRanksValidateScalarValidateSplitDimValidateSquareMatrixValidateUnbindDimValueOptionsvar_var_meanvdotviewview_as_complexview_as_realvmapvsplitvstackWebGPUKernelConfigWebGPUOnlyResultWebGPUTensorDatawhereWindowOptionsxlogyzeroszeros_like
torch.js· 2026
LegalTerms of UsePrivacy Policy
/
/
  1. docs
  3. torch.js
  5. torch
  7. nn
  9. functional

torch.nn.functional

Functions

  • layer_norm - Apply Layer Normalization.
  • rms_norm - RMS Normalization: normalizes by root mean square of activations (LLaMA, modern LLMs).
  • normalize - Lp Normalization: scales vectors to unit Lp norm along a dimension.
  • group_norm - Group Normalization: divides channels into groups and normalizes each group independently.
  • instance_norm - Instance Normalization: normalizes each channel independently for each sample.
  • batch_norm - Batch Normalization: normalize activations using batch statistics during training, running statistic
  • local_response_norm - Apply local response normalization.
  • log_softmax - Log-Softmax activation function: numerically stable log(softmax(x)).
  • nll_loss - Negative Log Likelihood (NLL) loss: standard loss for classification with pre-computed log-probabili
  • cross_entropy - Cross Entropy Loss: standard loss function for multi-class classification from raw logits.
  • mse_loss - Mean Squared Error (MSE) Loss: standard regression loss function.
  • l1_loss - L1 Loss (Mean Absolute Error): robust regression loss function.
  • dropout - Dropout regularization: randomly zeroes elements during training to prevent overfitting.
  • alpha_dropout - Alpha Dropout: dropout for self-normalizing neural networks (SNNs) with SELU.
  • feature_alpha_dropout - Randomly masks out entire channels with SELU-compatible dropout.
  • dropout1d - Randomly zeros out entire channels in 1D input (sequences).
  • dropout2d - Randomly zeros out entire channels in 2D input (spatial feature maps).
  • dropout3d - Randomly zeros out entire channels in 3D input (volumetric feature maps).
  • max_pool2d - 2D Max Pooling: downsamples feature maps by taking maximum values.
  • avg_pool2d - 2D Average Pooling: downsamples feature maps by averaging values.
  • max_pool3d - 3D Max Pooling: downsamples volumetric data by taking maximum values.
  • avg_pool3d - 3D Average Pooling: downsamples volumetric data by averaging values.
  • adaptive_avg_pool2d - 2D Adaptive Average Pooling: averages to fixed spatial size automatically.
  • adaptive_max_pool2d - 2D Adaptive Max Pooling: takes max values to fixed spatial size automatically.
  • adaptive_avg_pool3d - 3D Adaptive Average Pooling: averages to fixed volumetric size automatically.
  • adaptive_max_pool3d - 3D Adaptive Max Pooling: takes max values to fixed volumetric size automatically.
  • max_pool1d - 1D Max Pooling: downsamples sequences by taking maximum values.
  • avg_pool1d - 1D Average Pooling: downsamples sequences by averaging values.
  • adaptive_avg_pool1d - 1D Adaptive Average Pooling: averages to fixed output size automatically.
  • adaptive_max_pool1d - 1D Adaptive Max Pooling: takes max values to fixed output size automatically.
  • lp_pool1d - Apply 1D LP pooling.
  • lp_pool2d - Apply 2D LP pooling.
  • lp_pool3d - Apply 3D LP pooling.
  • max_pool1d_with_indices - Apply 1D max pooling with indices.
  • max_pool2d_with_indices - Applies 2D max pooling over an input signal and returns both the pooled values and their indices.
  • max_pool3d_with_indices - Apply 3D max pooling with indices.
  • max_unpool1d - Apply 1D max unpooling.
  • max_unpool2d - Apply 2D max unpooling.
  • max_unpool3d - Apply 3D max unpooling.
  • fractional_max_pool2d - Apply 2D fractional max pooling over an input signal.
  • fractional_max_pool3d - Apply 3D fractional max pooling over an input signal.
  • fractional_max_pool2d_with_indices - Apply 2D fractional max pooling and return both output and indices.
  • fractional_max_pool3d_with_indices - Apply 3D fractional max pooling and return both output and indices.
  • conv2d - 2D Convolution: applies learned filters to extract spatial features from images.
  • conv3d - Applies a 3D convolution over an input signal composed of several input planes.
  • conv_transpose1d - 1D transposed convolution: applies a 1D transposed convolution operator over an input signal.
  • conv_transpose2d - 2D Transposed Convolution (Deconvolution): upsamples spatial dimensions with learned parameters.
  • conv_transpose3d - 3D transposed convolution: applies a 3D transposed convolution operator over an input volume.
  • embedding - Generate embeddings by looking up indices in a weight matrix (dense vector representation table).
  • embedding_bag - Compute sums, means, or maxes of bags of embeddings.
  • one_hot - Convert class indices to one-hot encoded vectors (categorical representation).
  • binary_cross_entropy - Binary cross entropy loss.
  • binary_cross_entropy_with_logits - Binary cross entropy with logits loss.
  • smooth_l1_loss - Smooth L1 Loss: hybrid regression loss combining benefits of L1 and MSE.
  • huber_loss - Huber loss: robust regression loss with smooth L2 near zero, linear for outliers.
  • kl_div - Kullback-Leibler Divergence Loss: measures dissimilarity between probability distributions.
  • cosine_embedding_loss - Cosine Embedding Loss: learns similarity relationships using angular distance.
  • triplet_margin_loss - Triplet Margin Loss: learns embeddings with relative distances between triplets.
  • margin_ranking_loss - Margin ranking loss for learning relative ordering between pairs of samples.
  • hinge_embedding_loss - Hinge embedding loss for learning embeddings with margin constraints.
  • soft_margin_loss - Soft margin loss for binary classification with logistic regression.
  • unfold - Unfold: extracts sliding local blocks from image-like tensors (im2col operation).
  • fold - Combines sliding local blocks into a large tensor (col2im, inverse of unfold).
  • conv1d - 1D Convolution: applies learned filters to extract temporal or sequential features.
  • pad - Pad a tensor with various padding modes: constant fill, reflection, replication, or circular wrappin
  • interpolate - Resample (upsample or downsample) spatial dimensions of a tensor to new sizes using interpolation.
  • ctc_loss - CTC (Connectionist Temporal Classification) loss.
  • poisson_nll_loss - Poisson negative log likelihood loss for count data and event prediction.
  • gaussian_nll_loss - Gaussian (normal distribution) negative log likelihood loss for continuous predictions.
  • multilabel_margin_loss - Multi-label margin loss for ranking multiple positive classes with margin separation.
  • multilabel_soft_margin_loss - Multi-label soft margin loss using logistic loss for each class independently.
  • multi_margin_loss - Multi-class margin loss function for classification with custom margin.
  • triplet_margin_with_distance_loss - Triplet margin loss with custom distance function for flexible metric learning.
  • pixel_shuffle - Pixel Shuffle: rearranges channels into spatial dimensions for super-resolution upsampling.
  • pixel_unshuffle - Pixel Unshuffle: inverse of pixel shuffle, reorganizing channels into spatial dimensions.
  • channel_shuffle - Channel Shuffle: rearranges channels by dividing into groups and shuffling order.
  • pairwise_distance - Pairwise Distance: computes Lp distance between corresponding vector pairs.
  • cosine_similarity - Cosine Similarity: measures angular distance between vectors, invariant to magnitude.
  • pdist - Pairwise Distance: computes Lp distances between all pairs of vectors in a batch.
  • upsample - Upsample: increases spatial dimensions via interpolation (alias for interpolate).
  • upsample_nearest - Upsamples the input tensor using nearest neighbor interpolation.
  • upsample_bilinear - Upsamples the input tensor using bilinear interpolation.
  • grid_sample - Spatial Transformer Grid Sampling: samples input using learned spatial transformation grids.
  • affine_grid - Affine Grid Generation: converts affine transformation matrices to coordinate grids for grid_sample.
  • adaptive_max_pool1d_with_indices - Applies 1D adaptive max pooling and returns both output and indices.
  • adaptive_max_pool2d_with_indices - Applies 2D adaptive max pooling and returns both output and indices.
  • adaptive_max_pool3d_with_indices - Applies 3D adaptive max pooling and returns both output and indices.
  • multi_head_attention_forward - Multi-head attention forward pass.
  • grouped_mm - Performs grouped (multi-headed) matrix multiplication with optional bias and dtype casting.
  • scaled_grouped_mm - Performs batched (grouped) matrix multiplication with per-batch scaling.
  • scaled_mm - Performs scaled matrix multiplication with optional quantization and mixed-precision support.
  • linear - Applies a linear transformation: y = x @ weight.T + bias

Types

  • PoolWithIndicesResult - Result type for pooling operations that return both values and indices.
  • ReluFunctionalOptions - Options for relu functional operation.
  • RreluFunctionalOptions - Options for rrelu functional operation.
  • LeakyReluFunctionalOptions - Options for leaky_relu functional operation.
  • EluFunctionalOptions - Options for elu functional operation.
  • GluFunctionalOptions - Options for glu functional operation.
  • HardtanhFunctionalOptions - Options for hardtanh functional operation.
  • CeluFunctionalOptions - Options for celu functional operation.
  • HardshrinkFunctionalOptions - Options for hardshrink functional operation.
  • SoftshrinkFunctionalOptions - Options for softshrink functional operation.
  • SoftplusFunctionalOptions - Options for softplus functional operation.
  • SoftminFunctionalOptions - Options for softmin functional operation.
  • ScaledDotProductAttentionFunctionalOptions - Options for scaled_dot_product_attention functional operation.
  • MultiHeadAttentionFunctionalOptions - Options for multi_head_attention_forward functional operation.
  • UnfoldFunctionalOptions - Options for unfold functional operation.
  • FoldFunctionalOptions - Options for fold functional operation.
  • PadFunctionalOptions - Options for pad functional operation.
  • PairwiseDistanceFunctionalOptions - Options for pairwise_distance functional operation.
  • CosineSimilarityFunctionalOptions - Options for cosine_similarity functional operation.
  • PdistFunctionalOptions - Options for pdist functional operation.
  • NormalizeFunctionalOptions - Options for normalize functional operation.
  • LocalResponseNormFunctionalOptions - Options for local_response_norm functional operation.
  • BatchNormFunctionalOptions - Options for batch_norm functional operation.
  • LayerNormFunctionalOptions - Options for layer_norm functional operation.
  • RmsNormFunctionalOptions - Options for rms_norm functional operation.
  • GroupNormFunctionalOptions - Options for group_norm functional operation.
  • InstanceNormFunctionalOptions - Options for instance_norm functional operation.
  • NllLossFunctionalOptions - Options for nll_loss functional operation.
  • CrossEntropyFunctionalOptions - Options for cross_entropy functional operation.
  • MseLossFunctionalOptions - Options for mse_loss functional operation.
  • L1LossFunctionalOptions - Options for l1_loss functional operation.
  • BinaryCrossEntropyFunctionalOptions - Options for binary_cross_entropy functional operation.
  • BinaryCrossEntropyWithLogitsFunctionalOptions - Options for binary_cross_entropy_with_logits functional operation.
  • SmoothL1LossFunctionalOptions - Options for smooth_l1_loss functional operation.
  • HuberLossFunctionalOptions - Options for huber_loss functional operation.
  • KlDivFunctionalOptions - Options for kl_div functional operation.
  • CosineEmbeddingLossFunctionalOptions - Options for cosine_embedding_loss functional operation.
  • TripletMarginLossFunctionalOptions - Options for triplet_margin_loss functional operation.
  • MarginRankingLossFunctionalOptions - Options for margin_ranking_loss functional operation.
  • HingeEmbeddingLossFunctionalOptions - Options for hinge_embedding_loss functional operation.
  • SoftMarginLossFunctionalOptions - Options for soft_margin_loss functional operation.
  • DropoutFunctionalOptions - Options for dropout functional operations.
  • AlphaDropoutFunctionalOptions - Options for alpha_dropout and feature_alpha_dropout functional operations.
  • Conv1dFunctionalOptions - Options for 1D convolution functional operation.
  • Conv2dFunctionalOptions - Options for 2D convolution functional operation.
  • Conv3dFunctionalOptions - Options for 3D convolution functional operation.
  • ConvTranspose1dFunctionalOptions - Options for 1D transposed convolution functional operation.
  • ConvTranspose2dFunctionalOptions - Options for 2D transposed convolution functional operation.
  • ConvTranspose3dFunctionalOptions - Options for 3D transposed convolution functional operation.
  • MaxPool1dFunctionalOptions - Options for max_pool1d functional operation.
  • MaxPool2dFunctionalOptions - Options for max_pool2d functional operation.
  • MaxPool3dFunctionalOptions - Options for max_pool3d functional operation.
  • AvgPool1dFunctionalOptions - Options for avg_pool1d functional operation.
  • AvgPool2dFunctionalOptions - Options for avg_pool2d functional operation.
  • AvgPool3dFunctionalOptions - Options for avg_pool3d functional operation.
  • AdaptiveMaxPoolFunctionalOptions - Options for adaptive_max_pool*d functional operations.
  • LPPoolFunctionalOptions - Options for lp_pool*d functional operations.
  • MaxUnpoolFunctionalOptions - Options for max_unpool*d functional operations.
  • FractionalMaxPoolFunctionalOptions - Options for fractional_max_pool*d functional operations.
  • EmbeddingFunctionalOptions - Options for embedding functional operation.
  • EmbeddingBagFunctionalOptions - Options for embedding_bag functional operation.
  • InterpolateFunctionalOptions - Options for interpolate functional operation.
  • GridSampleFunctionalOptions - Options for grid_sample functional operation.
  • AffineGridFunctionalOptions - Options for affine_grid functional operation.
  • GroupedMMFunctionalOptions - Options for grouped_mm functional operation.
  • ScaledGroupedMMFunctionalOptions - Options for scaled_grouped_mm functional operation.
  • ScaledMMFunctionalOptions - Options for scaled_mm functional operation.
  • SoftmaxOptions -
  • KLDivOptions - Options for kl_div loss.
  • CTCLossOptions -
  • UpsampleOptions - Options for upsample operations.
  • UpsampleNearestOptions - Options for upsample_nearest.
  • UpsampleBilinearOptions - Options for upsample_bilinear.