ml-switcheroo ππ¦ΒΆ
A Deterministic, Specification-Driven Transpiler for Deep Learning Frameworks.
Interactive DemoΒΆ
Try it right in your browser! The engine runs entirely client-side using WebAssembly.
Live Transpiler Demo
Client-side WebAssembly EngineInitialize the engine to translate code between frameworks locally.
%%{init: {'flowchart': {'rankSpacing': 50, 'nodeSpacing': 20, 'padding': 35}}}%%
flowchart TD
%% --- 1. Font & Node Styling ---
%% Level 0: Red (Representations)
classDef l0Node fill: #ea4335, stroke: #ff7daf, stroke-width: 2px, color: white, font-family: 'Google Sans Normal', font-size: 16px, rx: 5px, ry: 5px;
%% Level 1: Blue (Frameworks)
classDef l1Node fill: #4285f4, stroke: #57caff, stroke-width: 2px, color: white, font-family: 'Google Sans Normal', font-size: 16px, rx: 5px, ry: 5px;
%% Level 2: Green (Numerical)
classDef l2Node fill: #34a853, stroke: #5cdb6d, stroke-width: 2px, color: white, font-family: 'Google Sans Normal', font-size: 16px, rx: 5px, ry: 5px;
%% Level 3: Yellow (Intermediate)
classDef l3Node fill: #f9ab00, stroke: #ffd427, stroke-width: 2px, color: white, font-family: 'Google Sans Normal', font-size: 16px, rx: 5px, ry: 5px;
%% Hardware: Navy (SASS) - Roboto Mono
classDef asmNode fill: #20344b, stroke: #57caff, stroke-width: 2px, color: white, font-family: 'Roboto Mono Normal', font-size: 14px, rx: 2px, ry: 2px;
%% --- 2. Subgraph Styling ---
%% White backgrounds to ensure text readability + visual grouping
classDef containerL0 fill: white, stroke: #ea4335, stroke-width: 3px, color: #20344b, font-family: 'Google Sans Medium', font-size: 20px;
classDef containerL1 fill: white, stroke: #4285f4, stroke-width: 3px, color: #20344b, font-family: 'Google Sans Medium', font-size: 20px;
classDef containerL2 fill: white, stroke: #34a853, stroke-width: 3px, color: #20344b, font-family: 'Google Sans Medium', font-size: 20px;
classDef containerL3 fill: white, stroke: #f9ab00, stroke-width: 3px, color: #20344b, font-family: 'Google Sans Medium', font-size: 20px;
classDef containerHW fill: white, stroke: #20344b, stroke-width: 3px, color: #20344b, font-family: 'Google Sans Medium', font-size: 20px;
%% --- 3. Diagram Structure ---
subgraph L0 [Level 0: Representations]
direction LR
HTML
TikZ
LaTeX
end
subgraph L1 [Level 1: High-Level]
direction LR
PyTorch
MLX
TensorFlow
Keras
FlaxNNX[Flax NNX]
Pax
end
subgraph L2 [Level 2: Numeric only]
direction LR
JAX
NumPy
end
subgraph L3 [Level 3: Standard IR]
direction LR
StableHLO[Stable HLO]
MLIR
end
subgraph LBottom [Level 4: ASM]
direction LR
SASS[NVIDIA SASS]
RDNA[AMD RDNA]
end
%% --- 4. Connections ---
TikZ ~~~ TensorFlow
TensorFlow ~~~ JAX
JAX ~~~ StableHLO
StableHLO ~~~ SASS
%% --- 5. Apply Styles ---
class HTML,TikZ,LaTeX l0Node;
class PyTorch,MLX,TensorFlow,Keras,FlaxNNX,Pax l1Node;
class JAX,NumPy l2Node;
class StableHLO,MLIR l3Node;
class SASS asmNode;
class RDNA asmNode;
class L0 containerL0;
class L1 containerL1;
class L2 containerL2;
class L3 containerL3;
class LBottom containerHW;
Documentation
Reference
- API Reference
- Operation Reference
- A
- ALiBi
- AMax
- AMin
- Abs
- Abs
- Absolute
- AcceleratorError
- AccumulateN
- Acos
- Acos
- Acosh
- Acosh
- ActivationLayer
- ActivityRegularization
- Adadelta
- Adafactor
- Adagrad
- Adam
- AdamW
- Adamax
- AdaptedTransformerFeedForward
- AdaptiveAveragePooling1D
- AdaptiveAveragePooling2D
- AdaptiveAveragePooling3D
- AdaptiveAvgPool1d
- AdaptiveGradClipState
- AdaptiveMaxPool1d
- AdaptiveMaxPooling1D
- AdaptiveMaxPooling2D
- AdaptiveMaxPooling3D
- Add
- AddDecayedWeightsState
- AddLayer
- AddN
- AddNoiseState
- Addbmm
- Addcdiv
- Addcmul
- AdditiveAttention
- Addmm
- Addmv
- Addmv
- Addr
- Adjoint
- AffineGridGenerator
- AliasCopy
- AlignTensors
- All
- AllClose
- AllToShardedLinear
- AlphaDropout
- AlphaDropout
- Aminmax
- Angle
- Any
- Append
- ApplyAlongAxis
- ApplyIfFiniteState
- ApplyOverAxes
- ApproxMaxK
- ApproxMinK
- Arange
- Arccos
- Arccos
- Arccosh
- Arccosh
- Arcsin
- Arcsin
- Arcsinh
- Arcsinh
- Arctan
- Arctan2
- Arctan
- Arctanh
- AreDeterministicAlgorithmsEnabled
- ArgMax
- ArgMin
- ArgPartition
- ArgSort
- ArgWhere
- Array
- Array2String
- ArrayAt
- ArrayEqual
- ArrayEquiv
- ArrayIterator
- ArrayLike
- ArrayRepr
- ArraySplit
- ArrayStr
- AsAnyArray
- AsArray
- AsArrayChkFinite
- AsContiguousArray
- AsFortranArray
- AsMatrix
- AsStrided
- AsTensor
- Asin
- Asinh
- AssertClose
- Astype
- AsyncEval
- AtLeast1D
- AtLeast2D
- AtLeast3D
- Atan
- Atan2
- Atanh
- Atom
- Attention
- AttentionLayer
- AttentionProjection
- AugMix
- AutoContrast
- Autocast
- AutodiffCheckpointType
- AuxOutput
- Average
- AverageGradients
- AverageLayer
- AveragePooling1D
- AveragePooling2D
- AveragePooling3D
- Avg
- AvgPool
- AvgPool1D
- AvgPool2D
- AvgPool3D
- B
- BFloat16
- BFloat16Tensor
- BLOCK_M
- Baddbmm
- Bartlett
- BartlettWindow
- BaseActivation
- BaseNormalization
- BatchNorm
- BatchNormBackwardElemt
- BatchNormBackwardReduce
- BatchNormElemt
- BatchNormGatherStats
- BatchNormGatherStatsWithCounts
- BatchNormStats
- BatchNormUpdateStats
- BatchNormalization
- BatchStat
- Bernoulli
- BesselI0
- BesselI0e
- BesselI1
- BesselI1e
- Betainc
- Bias
- Bidirectional
- Bilinear
- BinaryCrossEntropyWithLogits
- BinaryCrossentropy
- BinaryFocalCrossentropy
- Bincount
- Binomial
- Bit
- BitCast
- Bits16
- Bits1x8
- Bits2x4
- Bits4x2
- Bits8
- BitwiseAnd
- BitwiseCount
- BitwiseInvert
- BitwiseLeftShift
- BitwiseNot
- BitwiseOr
- BitwiseRightShift
- BitwiseXor
- Blackman
- BlackmanWindow
- Block
- BlockDiag
- BlockMaskedMM
- Bmm
- Bool
- BoolScalar
- BoolTensor
- BoolType
- BroadcastArrays
- BroadcastLike
- BroadcastShapes
- BroadcastTensors
- BroadcastTo
- Bucketize
- BuildMatrix
- BusDayCalendar
- BusDayCount
- BusDayOffset
- ByteTensor
- ByteType
- BytesType
- C
- CELU
- CTCLoss
- Cache
- CachedPartial
- Call
- CanCast
- Capsule
- Carry
- CartesianProd
- Cast
- CastBFloat16
- CastBool
- CastByte
- CastChar
- CastDouble
- CastFloat
- CastHalf
- CastInt
- CastInt64
- CastLong
- CastShort
- CategoricalCrossentropy
- CategoricalFocalCrossentropy
- CategoricalHinge
- CategoryEncoding
- CausalDepthwiseConv1D
- CausalMask
- Cbrt
- Cdist
- Ceil
- CeilInplace
- CeluInplace
- CenterCrop
- ChainMatmul
- CharTensor
- CharacterType
- Checkify
- Checkpoint
- Cholesky
- CholeskySolve
- Choose
- Chunk
- CifgLstmCellSimple
- CircleLoss
- Clamp
- ClampInplace
- ClampMax
- ClampMaxInplace
- ClampMin
- ClearCache
- Clip
- ClipByGlobalNormState
- ClipGradNorm
- ClipState
- Clip
- Clone
- Code
- ColumnConcatenation
- ColumnStack
- CombineMasks
- CommonType
- Compile
- Complex
- Complex128
- Complex64
- ComplexDouble
- ComplexFloating
- ComplexLongDouble
- ComplexWarning
- Compress
- ComputeCv
- ComputeMoments
- Concat
- Concatenate
- ConcatenateLayer
- ConcreteModuleType
- ConcreteModuleTypeBuilder
- Cond
- ConditionallyMaskState
- ConditionallyTransformState
- Conformer
- ConfusionMatrix
- Conj
- ConjPhysical
- Conjugate
- Constant
- ConstantPadNd
- ConstructFloat16
- ConstructFloat32
- ConstructFloat64
- Contiguous
- ControlDependency
- ControlVariate
- Conv
- Conv1D
- Conv1DTranspose
- Conv2D
- Conv2DTranspose
- Conv3DTranspose
- Conv3d
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- ConvT
- ConvTranspose
- ConvTranspose1d
- ConvTranspose2d
- ConvTranspose3d
- Convolution1D
- Convolution1DTranspose
- Convolution2D
- Convolution2DTranspose
- Convolution3D
- Convolution3DTranspose
- Convolve
- Copy
- CopyTo
- Copysign
- Corrcoef
- Correlate
- Cos
- Cos
- Cosh
- Cosh
- CosineAnnealingLR
- CosineEmbeddingLoss
- CosineSimilarity
- CosineSimilarityLoss
- CountNonzero
- Cov
- Cropping1D
- Cropping2D
- Cropping3D
- Cross
- CrossEntropyLoss
- Csingle
- CubeRoot
- CubedReLU
- CudaAvailable
- CumSum
- Cummax
- Cummin
- Cumprod
- CumulativeProd
- CumulativeSum
- CumulativeTrapezoid
- CurrentUpdateContext
- CustomAutograd
- CustomVjp
- CutMix
- CvExpectedValue
- CvState
- DSplit
- DStack
- Data
- DataLoader
- DataParallel
- Dataclass
- DeepCopyMemoTable
- Deg2Rad
- Deg2Rad
- Degrees
- Delete
- Dense
- DepthwiseConv1D
- DepthwiseConv2D
- Dequantize
- DeserializationStorageContext
- DeserializeLoss
- DeserializeOptimizer
- Det
- Detach
- DetachCopy
- Detach
- Device
- DeviceCpu
- DeviceObjType
- Diag
- DiagEmbed
- DiagIndices
- DiagIndicesFrom
- Diagflat
- Diagonal
- DiagonalCopy
- DiagonalScatter
- DiceLoss
- DictType
- Diff
- Digamma
- Digitize
- DisableCompile
- DisableTorchFunction
- DisableTorchFunctionSubclass
- Discretization
- DispatchKey
- DispatchKeySet
- Display
- Dist
- DistributedSum
- Div
- DivMod
- Divide
- DivideNoNan
- Dot
- DotList
- DotProductAttention
- Double
- DoubleStorage
- DoubleTensor
- Dropout
- Dropout2d
- Dropout3d
- Dropout
- Dsmm
- Dtype
- DtypeObject
- E
- ELU
- Ediff1d
- Eig
- Einsum
- EinsumDense
- EinsumLayer
- EinsumPath
- EmaState
- Embedding
- EmbeddingBag
- Empty
- EmptyLike
- EnableGrad
- EnumType
- Equal
- EqualAggregate
- EqualElementwise
- Equalization
- Erf
- ErfInv
- Erf
- Erfc
- ErfcInv
- Erfc
- ErrorReport
- EuclideanNorm
- EulerGamma
- Eval
- EvalMode
- EvalShape
- Event
- ExcludeDispatchKeyGuard
- ExecutionPlan
- Exp
- Exp2
- Exp2
- Exp
- ExpandCopy
- ExpandDims
- ExperimentalEnableNumpyBehavior
- Expm1
- Expm1
- ExportFunction
- ExportToDot
- Extract
- Eye
- FInfo
- FRnn
- Fabs
- FakeQuantizePerChannelAffine
- False
- FatalError
- FbgemmPackGemmMatrixFp16
- FbgemmPackQuantizedMatrix
- FeatureAlphaDropout
- FeatureAlphaDropout
- FeatureDropout
- FeatureDropout
- FeedForward
- FileCheck
- Fill
- FillDiagonal
- Fill
- FilterAll
- FilterAny
- FilterState
- FindDuplicates
- Fix
- Fix
- Flatnonsense
- Flatnonzero
- Flatten
- FlaxLayer
- Flexible
- Flip
- Fliplr
- Flipud
- Float
- Float16
- Float32
- Float4E2m1fnX2
- Float4_e2m1fn
- Float64
- Float8E4m3fn
- Float8E4m3fnuz
- Float8E5m2
- Float8E5m2fnuz
- Float8E8m0fnu
- Float8_e3m4
- Float8_e4m3
- Float8_e4m3b11fnuz
- Float8_e4m3fn
- Float8_e4m3fnuz
- Float8_e5m2
- Float8_e5m2fnuz
- Float8_e8m0fnu
- FloatPower
- FloatStorage
- FloatTensor
- FloatType
- Float
- Floating
- Floor
- FloorDivide
- FloorMod
- Floor
- Fmax
- Fmin
- Fmod
- ForiLoop
- ForkRngs
- FormatFloatPositional
- FormatFloatScientific
- Frac
- Frac
- Frexp
- FrobeniusNorm
- FromBuffer
- FromDlpack
- FromFile
- FromFlatState
- FromFunction
- FromIter
- FromNumpy
- FromPyFunc
- FromString
- FromTree
- Ftrl
- Full
- FullLike
- FullSoftmax
- FunctionSchema
- FusedMultiplyAdd
- Future
- FutureType
- GELU
- GLU
- GRU
- GRUCell
- GShardSharedEmbeddingSoftmax
- Gather
- GatherMM
- GaussianDropout
- GaussianNoise
- Gaussiannll
- Gcd
- Gcd
- Ge
- GeluApprox
- GeluFastApprox
- GenFromTxt
- GeneralConv
- Generator
- Generic
- GeomSpace
- Ger
- GetAbstractModel
- GetActiveMemory
- GetBufSize
- GetCacheMemory
- GetDefaultDevice
- GetDefaultDtype
- GetDefaultStream
- GetDevice
- GetErr
- GetErrCall
- GetFilePath
- GetFloat32MatmulPrecision
- GetInclude
- GetLoss
- GetNamedSharding
- GetNumInteropThreads
- GetNumThreads
- GetOptimizer
- GetPartitionSpec
- GetPeakMemory
- GetPrintOptions
- GetRngState
- GetState
- GlobalAveragePooling1D
- GlobalAveragePooling2D
- GlobalAveragePooling3D
- GlobalAvgPool1D
- GlobalAvgPool2D
- GlobalAvgPool3D
- GlobalMaxPool1D
- GlobalMaxPool2D
- GlobalMaxPool3D
- GlobalMaxPooling1D
- GlobalMaxPooling2D
- GlobalMaxPooling3D
- GlobalPooling
- Gpu
- Grad
- GradScaler
- Gradient
- Graph
- GraphExecutorState
- Graphdef
- Grayscale
- Greater
- GreaterEqual
- GridSampler
- GridSampler2d
- GridSampler3d
- Group
- GroupNorm
- GroupNormalization
- GroupQueryAttention
- GroupedQueryAttention
- Gt
- HSplit
- HStack
- Half
- Hamming
- HammingWindow
- HannWindow
- Hanning
- HardShrink
- HardSigmoid
- HardSilu
- HardSwish
- HardTanh
- HasLapack
- HasMkl
- HasOpenmp
- HasSpectral
- HashTensor
- HashedCrossing
- Hashing
- Heaviside
- HingeEmbeddingLoss
- HingeLoss
- Histc
- Histogram
- Histogram2d
- HistogramBinEdges
- HistogramDd
- Hsmm
- Hspmm
- HuberLoss
- Hypot
- I0
- I0
- IInfo
- INTEGER
- Identity
- IdentityNorm
- Igamma
- Igammac
- Imag
- ImportFunction
- ImportIrModule
- InTopK
- IndexAdd
- IndexCopy
- IndexExp
- IndexExpression
- IndexFill
- IndexPut
- IndexPut
- IndexReduce
- IndexSelect
- Indices
- Inexact
- Inf
- InferenceMode
- Infinite
- Info
- InitNumThreads
- InitialSeed
- Initializer
- InjectHyperparamsState
- InjectStatefulHyperparamsState
- Inner
- Input
- InputLayer
- InputSpec
- Insert
- InstanceNorm
- Int
- Int16
- Int16Tensor
- Int32
- Int4
- Int64
- Int8
- IntTensor
- Int
- Intc
- IntegerLookup
- Interp
- Intersect1d
- Intp
- Inverse
- InverseShortTimeFourierTransform
- Invert
- InvertPermutation
- IsAnomalyCheckNanEnabled
- IsAnomalyEnabled
- IsAutocastCacheEnabled
- IsAutocastCpuEnabled
- IsAutocastEnabled
- IsAutocastIpuEnabled
- IsAutocastXlaEnabled
- IsAvailable
- IsBusDay
- IsClose
- IsComplex
- IsComplexObj
- IsConj
- IsDType
- IsData
- IsDeterministicAlgorithmsWarnOnlyEnabled
- IsDistributed
- IsFinite
- IsFloatingPoint
- IsFortran
- IsGradEnabled
- IsIn
- IsInf
- IsInference
- IsInferenceModeEnabled
- IsNaN
- IsNaT
- IsNeg
- IsNegInf
- IsNonDecreasing
- IsNonZero
- IsPosInf
- IsReal
- IsRealObj
- IsSameSize
- IsScalar
- IsSigned
- IsStorage
- IsStrictlyIncreasing
- IsSubDtype
- IsTensor
- IsVulkanAvailable
- IsWarnAlwaysEnabled
- IterChildren
- IterGraph
- IterModules
- Iterable
- Ix
- JVP
- Jagged
- JaxLayer
- Jit
- KLDivergence
- KLDivergenceLoss
- Kaiser
- KaiserWindow
- KlDiv
- Kron
- KthValue
- L1
- L2Normalize
- LBeta
- LOWER_RIGHT
- LSTM
- LSTMCell
- LU
- LUSolve
- Lamb
- Lambda
- LanguageModel
- LanguageModelContinuousBatching
- Layer
- LayerNorm
- LayerNormalization
- LayerNormalizedLstmCellSimple
- LayerWrapper
- Layout
- LazyConvTranspose1d
- LazyConvTranspose2d
- LazyConvTranspose3d
- Lcm
- Lcm
- LdExp
- Ldexp
- Le
- LeakyReLU
- LeftShift
- Lerp
- Less
- LessEqual
- LexSort
- Lgamma
- LinSpace
- Linear
- LinearGeneral
- LinearT
- Lion
- List
- ListAdd
- ListAverage
- ListConcatenate
- ListMaximum
- ListMinimum
- ListMultiply
- ListSubtract
- LittleEndian
- LoRA
- LoRALinear
- LoRAParam
- Load
- LoadTxt
- LobPCG
- Log
- Log10
- Log10
- Log1p
- Log1p
- Log2
- Log2
- LogAddExp
- LogAddExp2
- LogCoshLoss
- LogCumSumExp
- LogDet
- LogSigmoid
- LogSoftmax
- LogSpace
- LogSumExp
- Log
- Logcosh
- LogicalAnd
- LogicalAxisRules
- LogicalNot
- LogicalOr
- LogicalXor
- Logit
- Logit
- LongLong
- LongTensor
- Loss
- LossScaleOptimizer
- LstSq
- LstmCellSimple
- Lt
- LuUnpack
- M
- MA
- MGrid
- MLPBlock
- MSELoss
- MakeAttentionMask
- MakeCausalMask
- ManualSeed
- MapState
- MarginRankingLoss
- Marginranking
- MaskIndices
- MaskedFill
- MaskedNode
- MaskedScatter
- MaskedSelect
- MaskedState
- Masking
- MatMul
- Matrix
- MatrixExp
- MatrixPower
- MatrixRank
- MatrixTranspose
- Matvec
- Max
- MaxNumBoundingBoxes
- MaxPool
- MaxPool1d
- MaxPool1dWithIndices
- MaxPool2d
- MaxPool3d
- MaxPooling1D
- MaxPooling2D
- MaxPooling3D
- MaxUnpool1d
- MaxUnpool2d
- MaxUnpool3d
- Maximum
- MayShareMemory
- Mean
- MeanAbsoluteError
- MeanAbsolutePercentageError
- MeanSquaredError
- MeanSquaredLogarithmicError
- Median
- MelSpectrogram
- Memmap
- MemoryFormat
- Merge
- MergeContext
- MergeState
- MergeTypeFromTypeComment
- Meshgrid
- Metric
- Min
- MinPool
- MinScalarType
- MinTypeCode
- Minimum
- Mish
- MixUp
- MkldnnAdaptiveAvgPool2d
- MkldnnConvolution
- MkldnnLinearBackwardWeights
- MkldnnMaxPool2d
- MkldnnMaxPool3d
- MkldnnRnnLayer
- Mm
- Mod
- Mode
- ModelAndOptimizer
- Modf
- Module
- ModuleDict
- ModuleList
- MoveAxis
- Movedim
- Mse
- Msort
- Mul
- MultiHeadAttention
- MultiMetric
- MultiSteps
- MultiStepsState
- MultiTransformState
- Multinomial
- Multiply
- MultiplyNoNan
- Muon
- Mv
- Mvlgamma
- NONE
- Nadam
- Names
- Nan
- NanArgMax
- NanArgMin
- NanCumProd
- NanCumSum
- NanMax
- NanMean
- NanMedian
- NanMin
- NanPercentile
- NanProd
- NanQuantile
- NanStd
- NanSum
- NanToNum
- NanToNum
- NanVar
- Narrow
- NarrowCopy
- NativeBatchNorm
- NativeChannelShuffle
- NativeDropout
- NativeGroupNorm
- NativeLayerNorm
- NativeNorm
- NdArray
- NdEnumerate
- NdIndex
- NdIter
- Ndim
- Ndtri
- Negative
- NegativeAlias
- NegativeAlias
- Negative
- NestedIters
- NewAxis
- NextAfter
- Nll
- NoGrad
- Node
- NodeStates
- NonNegativeParamsState
- NonZero
- Nop
- Norm
- Normal
- Normalization
- Normalize
- Not
- NotEqual
- NotEqualAlias
- Nothing
- NuclearNorm
- Number
- Numel
- OGrid
- Object
- Object
- OfType
- OneHot
- Ones
- OnesLike
- OptArray
- OptState
- OptVariable
- OptaxTest
- OptimizedLSTMCell
- Optimizer
- Optional
- OrgQr
- OutOfMemoryError
- Outer
- PReLU
- PackBits
- Pad
- PairwiseDistance
- Param
- Parameter
- ParameterDict
- ParameterList
- Partition
- PartitionName
- PartitionState
- PathContains
- PerChannelSymmetric
- PerDimScale
- PerTensorAffine
- PerTensorSymmetric
- Percentile
- Permute
- PermuteCopy
- PermuteDims
- Perturbation
- Pi
- Piecewise
- Pinverse
- Pipeline
- PipelinedTransformer
- PixelShuffle
- PixelUnshuffle
- Place
- Pmap
- Poisson
- PoissonLoss
- PoissonNllLoss
- Polar
- Poly
- Poly1d
- PolyAdd
- PolyDer
- PolyDiv
- PolyFit
- PolyInt
- PolyMul
- PolySub
- PolyVal
- Polygamma
- Pool
- Pooling
- Pooling1D
- Pop
- Positive
- Pow
- Power
- PrepareMultiprocessingEnvironment
- PreserveFormat
- PrintOptions
- Prod
- ProfilerAllowCudagraphCuptiLazyReinitCuda12
- ProjectLastDim
- PromoteTypes
- Ptp
- PureState
- Put
- PutAlongAxis
- PutMask
- Pytree
- QInt32
- QInt8
- QPerChannelAxis
- QPerChannelScales
- QPerChannelZeroPoints
- QQLinear
- QScale
- QScheme
- QUInt2x4Storage
- QUInt4x2Storage
- QUInt8Storage
- QZeroPoint
- Qqmm
- Qr
- Quantile
- Quantize
- QuantizePerChannel
- QuantizePerTensor
- QuantizePerTensorDynamic
- QuantizedAllToShardedLinear
- QuantizedBatchNorm
- QuantizedEmbedding
- QuantizedGru
- QuantizedGruCell
- QuantizedLinear
- QuantizedLstm
- QuantizedLstmCell
- QuantizedMatmul
- QuantizedMaxPool1d
- QuantizedMaxPool2d
- QuantizedMaxPool3d
- QuantizedRnnReluCell
- QuantizedRnnTanhCell
- QuantizedShardedToAllLinear
- Quint2x4
- Quint4x2
- Quint8
- RMSNorm
- RMSNormalization
- RMSprop
- RNN
- RNNBase
- RNNCellBase
- RPC_AVAILABLE
- RRefType
- R
- Rad2Deg
- Rad2Deg
- Radians
- Rand
- RandAugment
- RandInt
- RandIntLike
- RandLike
- RandPerm
- Randn
- RandnLike
- RandomBrightness
- RandomColorDegeneration
- RandomColorJitter
- RandomContrast
- RandomCrop
- RandomElasticTransform
- RandomErasing
- RandomFlip
- RandomGaussianBlur
- RandomGrayscale
- RandomHorizontalFlip
- RandomHue
- RandomInvert
- RandomPerspective
- RandomPosterization
- RandomRotation
- RandomSaturation
- RandomSharpness
- RandomShear
- RandomTranslation
- RandomVerticalFlip
- RandomZoom
- Range
- Ravel
- RavelMultiIndex
- ReLU
- ReLU2
- ReLU6
- ReadVitals
- Real
- RealIfClose
- Reciprocal
- ReciprocalNoNan
- Reciprocal
- RecursiveMap
- ReduceMin
- ReduceProd
- ReduceStd
- ReduceSum
- ReduceVariance
- RegisterDataType
- Relu
- Remainder
- Remat
- Renorm
- Repeat
- RepeatInterleave
- RepeatVector
- ReplaceByPureDict
- ResNet
- Rescaling
- Reseed
- ResetPeakMemory
- Reshape
- Resize
- Resizing
- ResolveConj
- ResolveNeg
- ResultType
- ReversibleEmbedding
- RightShift
- Rint
- Rngs
- RoPE
- Roll
- RollAxis
- Roots
- Rot90
- Round
- Round
- RowStack
- Rsqrt
- Rsqrt
- Rsub
- SELU
- SGD
- STFT
- STFTSpectrogram
- STRING
- Saddmm
- Save
- SaveGguf
- SaveSafetensors
- SaveTxt
- SaveZ
- SaveZCompressed
- ScalarMul
- ScalarTensor
- ScalarType
- ScaleByTrustRatioState
- ScaleState
- Scan
- Scatter
- ScatterAdd
- ScatterReduce
- ScriptClass
- ScriptClassFunction
- ScriptDict
- ScriptDictIterator
- ScriptDictKeyIterator
- ScriptFunction
- ScriptList
- ScriptListIterator
- ScriptMethod
- ScriptModule
- ScriptModuleSerializer
- ScriptObject
- ScriptObjectProperty
- SeLU
- SearchSorted
- Seed
- SegmentMax
- SegmentMean
- SegmentMin
- SegmentProd
- SegmentReduce
- SegmentSum
- SegmentedMm
- Select
- SelectCopy
- SelectScatter
- SeparableConv1D
- SeparableConv2D
- SeparableConvolution1D
- SeparableConvolution2D
- Sequential
- SerializationStorageContext
- SerializeLoss
- SerializeOptimizer
- SetAnomalyEnabled
- SetAutocastDtype
- SetCacheLimit
- SetDefaultDevice
- SetDefaultDtype
- SetDefaultStream
- SetDeterministicDebugMode
- SetDiff1d
- SetErr
- SetErrCall
- SetFloat32MatmulPrecision
- SetFlushDenormal
- SetGradEnabled
- SetMemoryLimit
- SetMode
- SetNumInteropThreads
- SetNumThreads
- SetPrintOptions
- SetRngState
- SetVital
- SetWarnAlways
- SetWiredLimit
- SetXor1d
- Sgn
- Shape
- ShardMap
- ShardedToAllLinear
- SharesMemory
- Short
- ShortStorage
- ShouldSkipUpdateFunction
- ShowConfig
- SiLU
- Sigmoid
- SigmoidCrossEntropy
- Sigmoid
- Sign
- Signbit
- SignedInteger
- SimpleCell
- SimpleRNN
- SimpleRNNCell
- Sin
- Sin
- Sinc
- Sinc
- Single
- Sinh
- Sinh
- SinusoidalPositionalEncoding
- Size
- Slice
- SliceCopy
- SliceInverse
- SliceScatter
- SliceUpdate
- Slogdet
- Smm
- Smoothl1
- SnapshotState
- SobolSample
- SoftSign
- Softmax
- Softmax2d
- Softmin
- Softplus
- Softshrink
- Solarization
- Solve
- Sort
- SortComplex
- Spacing
- SparseBsc
- SparseBscTensor
- SparseBsr
- SparseBsrTensor
- SparseCategoricalCrossentropy
- SparseCompressed
- SparseCompressedTensor
- SparseCoo
- SparseCooTensor
- SparseCsc
- SparseCscTensor
- SparseCsr
- SparseCsrTensor
- Sparsemax
- SpatialDropout1D
- SpatialDropout2D
- SpatialDropout3D
- SpectralNorm
- SpectralNormalization
- Split
- SplitCopy
- SplitRngs
- SplitWithSizes
- Spmm
- Sqrt
- Sqrt
- Square
- Square
- SquaredDifference
- SquaredHinge
- SquaredHingeLoss
- Squeeze
- SqueezeCopy
- Sspaddmm
- Stack
- StackedRNNCells
- StackedTransformer
- Standardize
- State
- StaticModule
- Std
- StdMean
- Step
- StepActivation
- StepLR
- StochasticResidual
- StopGradient
- Storage
- StorageBase
- Stream
- StreamContext
- StreamObjType
- Strided
- StringLookup
- StringType
- Sub
- Subtract
- SubtractLayer
- Sum
- Svd
- SvdLowrank
- SwapAxes
- Swapdims
- Swish
- Switch
- SymBool
- SymBoolType
- SymConstrainRange
- SymConstrainRangeForSize
- SymFloat
- SymFreshSize
- SymInt
- SymIte
- SymMax
- SymMin
- SymNot
- SymSqrt
- SymSum
- Symeig
- Synchronize
- T
- TCopy
- TFSMLayer
- TUPLE
- TYPE_CHECKING
- T_module
- Tabulate
- Tag
- Take
- TakeAlongAxis
- TakeAlongDim
- Tan
- Tan
- Tanh
- Tanh
- Tanhshrink
- TemporalShifting
- Tensor
- TensorSplit
- TensorType
- Tensordot
- Test
- TextVectorization
- Threshold
- Threshold
- Tile
- TimeDistributed
- Timedelta64
- ToDlpack
- ToPureDict
- ToTensor
- TopK
- TopKValues
- TorchFunctional
- TorchLoad
- TorchModuleWrapper
- TorchSave
- Trace
- TraceState
- TrainMode
- TrainablePositionalEmbedding
- Transformer
- TransformerDecoderLayer
- TransformerEncoderDecoder
- TransformerEncoderLayer
- TransformerFeedForward
- TransformerFeedForwardMoe
- TransformerLm
- Transformerdecoder
- Transformerencoder
- Transpose
- Transpose2D
- TransposeCopy
- Trapezoid
- Trapz
- Tri
- TriangularSolve
- Tril
- TrilIndices
- TrilIndicesFrom
- TrimZeros
- TripletMarginLoss
- Triu
- TriuIndices
- TriuIndicesFrom
- TrueDiv
- TrueDivide
- True
- Trunc
- TruncInplace
- TverskyLoss
- Type
- TypeVarA
- TypeVarM
- Typecodes
- Typename
- UByte
- UFunc
- UInt
- UInt16
- UInt32
- UInt64
- UInt8
- UIntC
- UIntP
- ULong
- ULongLong
- UShort
- Uint1
- Uint2
- Uint3
- Uint4
- Uint5
- Uint6
- Uint7
- Unbind
- UnbindCopy
- Unflatten
- Unfold
- UnfoldCopy
- Unicode
- UnifyTypeList
- Union1d
- Unique
- UniqueAll
- UniqueConsecutive
- UniqueCounts
- UniqueInverse
- UniqueValues
- UnitNormalization
- UnpackBits
- UnravelIndex
- UnsafeChunk
- UnsafeSplit
- UnsafeSplitWithSizes
- UnsignedInteger
- UnsortedSegmentMax
- UnsortedSegmentMin
- UnsortedSegmentProd
- UnsortedSegmentSum
- Unsqueeze
- UnsqueezeCopy
- Unstack
- Unwrap
- UpSampling1D
- UpSampling2D
- UpSampling3D
- UpdateContext
- UpdateCvState
- UpdateState
- Upsample
- UseDeterministicAlgorithms
- UseEagerSharding
- V
- VMap
- VQNgrammer
- Value
- ValueAndGrad
- ValuesCopy
- Vander
- VanillaBlock
- VanillaNet
- Var
- VarMean
- Variable
- Variance
- Vdot
- Vecdot
- Vecmat
- VectorQuantization
- VectorQuantizer
- Vectorize
- View
- ViewAsComplex
- ViewAsComplexCopy
- ViewAsReal
- ViewAsRealCopy
- ViewCopy
- VisionTransformer
- VitEntryLayers
- VitExitLayers
- VitalsEnabled
- Vjp
- Vsplit
- Vstack
- Wait
- WeightNorm
- Where
- WhileLoop
- WithPartitioning
- WrappedSchedule
- Wrapper
- Xdivy
- Xlog1py
- Xlogy
- Xlogy
- ZeroFraction
- ZeroNansState
- ZeroPadding1D
- ZeroPadding2D
- ZeroPadding3D
- Zero
- Zeros
- ZerosLike
- Zeta
- __array_namespace_info__
- activation
- activation_relu_or_gelu
- adaptive_average_pool
- adaptive_grad_clip
- adaptive_max_pool
- add_decayed_weights
- add_noise
- add_scale
- addressable_data
- align_corners
- alpha
- apply
- apply_along_axis
- apply_gradients
- apply_if_finite
- apply_mask
- apply_over_axes
- approximate
- args
- array_equal
- array_equiv
- array_split
- as_string
- assert_equal
- associative_scan
- at
- atleast_1d
- atleast_2d
- atleast_3d
- attr
- average_pool
- axis
- base
- batch_normalization
- batch_shape
- beta
- bias_correction
- bias_hh
- bias_ih
- bias_k
- bias_v
- binary_cross_entropy
- binary_crossentropy
- binary_focal_crossentropy
- bits
- bitwise_and
- bitwise_invert
- bitwise_left_shift
- bitwise_not
- bitwise_or
- bitwise_right_shift
- bitwise_xor
- bool
- broadcast
- broadcast_arrays
- broadcast_shapes
- broadcast_to
- byte
- cached
- can_cast
- cast_like
- categorical
- categorical_crossentropy
- categorical_hinge
- cdouble
- ceil_mode
- chain
- char
- character
- children
- chisquare
- choice
- cholesky_inverse
- clear
- clip_by_block_rms
- clip_by_global_norm
- clip_grad_norm
- close
- column_stack
- complex
- compute_mask
- conditionally_mask
- conditionally_transform
- constant
- constant_schedule
- control_delta_method
- control_variates_jacobians
- conv1d_input
- conv1d_weight
- conv2d_input
- conv2d_weight
- conv3d_input
- conv3d_weight
- conv_transpose
- conv_transpose1d
- conv_transpose2d
- conv_transpose3d
- convert_to_numpy
- convert_to_tensor
- convex_kl_divergence
- copy_to_host_async
- cosine_decay_schedule
- cosine_distance
- cosine_onecycle_schedule
- cosine_similarity
- count
- count_nonzero
- cpu
- create_mask
- cross_entropy
- ctc_decode
- ctc_loss
- ctc_loss_with_forward_probs
- cumulative_prod
- cumulative_sum
- custom_gradient
- d_model
- data_ptr
- decoder
- depthwise_conv
- deserialize
- device_ids
- device_mesh
- device_type
- diag_indices
- diag_indices_from
- dilation
- dim
- dims
- dirichlet
- div_value
- divide_no_nan
- dot_product_attention
- downscale_factor
- dropout1
- dropout2
- dropout3
- eigh
- eigvals
- eigvalsh
- einsum_path
- ema
- embedding_dim
- empty_like
- enable_grad
- encoder
- end_dim
- eps
- erfcx
- estimate_control_variate_coefficients
- euler_gamma
- expand_dims
- expit
- exponential
- exponential_decay
- extend
- extract_sequences
- eye
- f
- fft
- fft2
- fftfreq
- fftn
- fftshift
- fill_diagonal
- filters
- find_duplicates
- flatten_parameters
- floor_divide
- fn
- fori_loop
- forward
- freeze
- from_dlpack
- full_like
- gamma
- gammainc
- gammaincc
- gammaln
- gaussian_nll_loss
- geometric
- get
- get_all_with_path
- get_item
- get_parameter
- get_printoptions
- glorot_normal
- glorot_uniform
- greater_equal
- group_size
- groups
- gumbel
- hard_shrink
- hard_sigmoid
- hard_silu
- hard_swish
- hard_tanh
- he_normal
- he_uniform
- head
- head_dim
- hfft
- hidden_size
- hinge_loss
- histogram_bin_edges
- huber_loss
- ifft
- ifft2
- ifftn
- ifftshift
- ihfft
- in1_features
- in2_features
- in_features
- in_top_k
- init
- inject_hyperparams
- inject_stateful_hyperparams
- inplace
- int2
- interpolate
- inv
- invert_permutation
- irfft
- irfft2
- irfftn
- is_tensor
- istft
- item
- items
- itemsize
- join
- join_schedules
- k
- kaiming_normal
- kaiming_uniform
- keep_params_nonnegative
- kernel_size
- key
- keys
- kl_divergence
- l1_loss
- l2_loss
- label_smoothing
- lambd
- laplace
- layer_norm
- layers
- leaky_relu
- left_shift
- less_equal
- linear1
- linear2
- linear_onecycle_schedule
- linear_schedule
- linear_to_mel_weight_matrix
- load_state_dict
- log_cosh
- log_ndtr
- log_prob
- log_sigmoid
- log_softmax
- logger
- logical_and
- logical_not
- logical_or
- logical_xor
- logistic
- lognormal
- long
- lower
- lu_factor
- map
- map_params
- margin
- margin_ranking_loss
- mask
- mask_check
- mask_indices
- mask_mod
- masked
- matrix_norm
- matrix_power
- matrix_rank
- matrix_transpose
- max_norm
- max_pool
- max_pool1d
- max_pool2d
- max_pool3d
- max_val
- measure_valued_estimation_mean
- measure_valued_estimation_std
- measure_valued_jacobians
- merge_masks
- metadata
- min_val
- modules
- moments
- momentum
- moving_avg_baseline
- mse_loss
- multi_dot
- multi_transform
- multivariate_normal
- n
- name
- named_chain
- named_modules
- named_parameters
- nan_to_num
- nbytes
- negative_slope
- nll_loss
- no_grad
- nonlinearity
- norm1
- norm2
- norm3
- norm_first
- norm_type
- not_equal
- ntxent
- num_embeddings
- num_features
- num_groups
- num_heads
- num_layers
- one_hot
- ones
- ones_like
- orthogonal
- orthogonal
- out_features
- out_proj
- output_device
- output_ratio
- output_size
- p
- padding
- padding_idx
- parameters
- parameters_to_vector
- params
- pareto
- pathwise_jacobians
- per_example_global_norm_clip
- per_example_layer_norm_clip
- permutation
- permute_dims
- piecewise_constant_schedule
- piecewise_interpolate_schedule
- pinv
- polynomial_schedule
- popitem
- predict
- promote_types
- psi
- psnr
- random_like
- rank
- ravel_multi_index
- rayleigh
- register_buffer
- register_parameter
- remove
- remove_axis
- replace
- reset
- reshape_weight_to_matrix
- result_type
- rfft
- rfft2
- rfftfreq
- rfftn
- right_shift
- rms_norm
- rms_normalization
- safe_softmax_cross_entropy
- saturate_cast
- savez_compressed
- scale
- scale_factor
- scale_grad_by_freq
- scaled_dot_product_attention
- scatter_update
- score_function_jacobians
- segment_max
- segment_sum
- selective_transform
- separable_conv
- serialize
- set
- set_printoptions
- sgdr_schedule
- shuffle
- sigmoid_binary_cross_entropy
- sigmoid_focal_loss
- skip_large_updates
- skip_not_finite
- slice_update
- smooth_l1_loss
- smooth_labels
- snapshot
- soft_shrink
- softmax_cross_entropy
- softmax_cross_entropy_with_integer_labels
- solve_triangular
- sparse
- sparse
- sparse_categorical_crossentropy
- sparse_plus
- sparse_sigmoid
- split_key_like
- squared_error
- squareplus
- start_dim
- state_dict
- static_graph
- stop_gradient
- stride
- strides
- svdvals
- swap
- take_along_axis
- tanh_shrink
- tensorinv
- tensorsolve
- test_import
- to_dense
- to_device
- tolist
- top_k
- train
- training
- tree_add_scalar_mul
- tree_l1_norm
- tree_l2_norm
- tree_linf_norm
- tree_map_params
- tree_scalar_mul
- triangular
- tril_indices
- tril_indices_from
- trim_zeros
- triu_indices
- triu_indices_from
- true_divide
- trunc_normal
- truncated_normal
- unflattened_size
- uniform
- unique_all
- unique_counts
- unique_inverse
- unique_values
- unravel_index
- unwrap_random_key_data
- update
- update_infinity_moment
- update_moment
- update_moment_per_elem_norm
- upper
- upsample_bilinear
- upsample_nearest
- upscale_factor
- value_and_grad
- values
- variant
- vector_norm
- vectorized_map
- view_as_complex
- view_as_real
- wald
- warmup_constant_schedule
- warmup_cosine_decay_schedule
- warmup_exponential_decay_schedule
- weight
- weight_hh
- weight_ih
- while_loop
- xavier_normal
- xavier_uniform
- zero_grad
- zero_nans
- zeros
- zeros_like