-
SOTA
-
Accelerator Toolkit
-
Deep Learning Toolkit
-
-
- Resume
- Add
- AlphaDropout
- AdditiveAttention
- Attention
- Average
- AvgPool1D
- AvgPool2D
- AvgPool3D
- BatchNormalization
- Bidirectional
- Concatenate
- Conv1D
- Conv1DTranspose
- Conv2D
- Conv2DTranspose
- Conv3D
- Conv3DTranspose
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Dense
- Cropping1D
- Cropping2D
- Cropping3D
- DepthwiseConv2D
- Dropout
- Embedding
- Flatten
- ELU
- Exponential
- GaussianDropout
- GaussianNoise
- GlobalAvgPool1D
- GlobalAvgPool2D
- GlobalAvgPool3D
- GlobalMaxPool1D
- GlobalMaxPool2D
- GlobalMaxPool3D
- GRU
- GELU
- Input
- LayerNormalization
- LSTM
- MaxPool1D
- MaxPool2D
- MaxPool3D
- MultiHeadAttention
- HardSigmoid
- LeakyReLU
- Linear
- Multiply
- Permute3D
- Reshape
- RNN
- PReLU
- ReLU
- SELU
- Output Predict
- Output Train
- SeparableConv1D
- SeparableConv2D
- SimpleRNN
- SpatialDropout
- Sigmoid
- SoftMax
- SoftPlus
- SoftSign
- Split
- UpSampling1D
- UpSampling2D
- UpSampling3D
- ZeroPadding1D
- ZeroPadding2D
- ZeroPadding3D
- Swish
- TanH
- ThresholdedReLU
- Substract
- Show All Articles (63) Collapse Articles
-
-
-
-
- Exp
- Identity
- Abs
- Acos
- Acosh
- ArgMax
- ArgMin
- Asin
- Asinh
- Atan
- Atanh
- AveragePool
- Bernouilli
- BitwiseNot
- BlackmanWindow
- Cast
- Ceil
- Celu
- ConcatFromSequence
- Cos
- Cosh
- DepthToSpace
- Det
- DynamicTimeWarping
- Erf
- EyeLike
- Flatten
- Floor
- GlobalAveragePool
- GlobalLpPool
- GlobalMaxPool
- HammingWindow
- HannWindow
- HardSwish
- HardMax
- lrfft
- lsNaN
- Log
- LogSoftmax
- LpNormalization
- LpPool
- LRN
- MeanVarianceNormalization
- MicrosoftGelu
- Mish
- Multinomial
- MurmurHash3
- Neg
- NhwcMaxPool
- NonZero
- Not
- OptionalGetElement
- OptionalHasElement
- QuickGelu
- RandomNormalLike
- RandomUniformLike
- RawConstantOfShape
- Reciprocal
- ReduceSumInteger
- RegexFullMatch
- Rfft
- Round
- SampleOp
- Shape
- SequenceLength
- Shrink
- Sin
- Sign
- Sinh
- Size
- SpaceToDepth
- Sqrt
- StringNormalizer
- Tan
- TfldfVectorizer
- Tokenizer
- Transpose
- UnfoldTensor
- lslnf
- ImageDecoder
- Inverse
- Show All Articles (65) Collapse Articles
-
-
-
- Add
- AffineGrid
- And
- BiasAdd
- BiasGelu
- BiasSoftmax
- BiasSplitGelu
- BitShift
- BitwiseAnd
- BitwiseOr
- BitwiseXor
- CastLike
- CDist
- CenterCropPad
- Clip
- Col2lm
- ComplexMul
- ComplexMulConj
- Compress
- ConvInteger
- Conv
- ConvTranspose
- ConvTransposeWithDynamicPads
- CropAndResize
- CumSum
- DeformConv
- DequantizeBFP
- DequantizeLinear
- DequantizeWithOrder
- DFT
- Div
- DynamicQuantizeMatMul
- Equal
- Expand
- ExpandDims
- FastGelu
- FusedConv
- FusedGemm
- FusedMatMul
- FusedMatMulActivation
- GatedRelativePositionBias
- Gather
- GatherElements
- GatherND
- Gemm
- GemmFastGelu
- GemmFloat8
- Greater
- GreaterOrEqual
- GreedySearch
- GridSample
- GroupNorm
- InstanceNormalization
- Less
- LessOrEqual
- LongformerAttention
- MatMul
- MatMulBnb4
- MatMulFpQ4
- MatMulInteger
- MatMulInteger16
- MatMulIntergerToFloat
- MatMulNBits
- MaxPoolWithMask
- MaxRoiPool
- MaxUnPool
- MelWeightMatrix
- MicrosoftDequantizeLinear
- MicrosoftGatherND
- MicrosoftGridSample
- MicrosoftPad
- MicrosoftQLinearConv
- MicrosoftQuantizeLinear
- MicrosoftRange
- MicrosoftTrilu
- Mod
- MoE
- Mul
- MulInteger
- NegativeLogLikelihoodLoss
- NGramRepeatBlock
- NhwcConv
- NhwcFusedConv
- NonMaxSuppression
- OneHot
- Or
- PackedAttention
- PackedMultiHeadAttention
- Pad
- Pow
- QGemm
- QLinearAdd
- QLinearAveragePool
- QLinearConcat
- QLinearConv
- QLinearGlobalAveragePool
- QLinearLeakyRelu
- QLinearMatMul
- QLinearMul
- QLinearReduceMean
- QLinearSigmoid
- QLinearSoftmax
- QLinearWhere
- QMoE
- QOrderedAttention
- QOrderedGelu
- QOrderedLayerNormalization
- QOrderedLongformerAttention
- QOrderedMatMul
- QuantizeLinear
- QuantizeWithOrder
- Range
- ReduceL1
- ReduceL2
- ReduceLogSum
- ReduceLogSumExp
- ReduceMax
- ReduceMean
- ReduceMin
- ReduceProd
- ReduceSum
- ReduceSumSquare
- RelativePositionBias
- Reshape
- Resize
- RestorePadding
- ReverseSequence
- RoiAlign
- RotaryEmbedding
- ScatterElements
- ScatterND
- SequenceAt
- SequenceErase
- SequenceInsert
- Sinh
- Slice
- SparseToDenseMatMul
- SplitToSequence
- Squeeze
- STFT
- StringConcat
- Sub
- Tile
- TorchEmbedding
- TransposeMatMul
- Trilu
- Unsqueeze
- Where
- WordConvEmbedding
- Xor
- Show All Articles (134) Collapse Articles
-
- Attention
- AttnLSTM
- BatchNormalization
- BiasDropout
- BifurcationDetector
- BitmaskBiasDropout
- BitmaskDropout
- DecoderAttention
- DecoderMaskedMultiHeadAttention
- DecoderMaskedSelfAttention
- Dropout
- DynamicQuantizeLinear
- DynamicQuantizeLSTM
- EmbedLayerNormalization
- GemmaRotaryEmbedding
- GroupQueryAttention
- GRU
- LayerNormalization
- LSTM
- MicrosoftMultiHeadAttention
- QAttention
- RemovePadding
- RNN
- Sampling
- SkipGroupNorm
- SkipLayerNormalization
- SkipSimplifiedLayerNormalization
- SoftmaxCrossEntropyLoss
- SparseAttention
- TopK
- WhisperBeamSearch
- Show All Articles (15) Collapse Articles
-
-
-
-
-
-
-
-
-
-
- AdditiveAttention
- Attention
- BatchNormalization
- Bidirectional
- Conv1D
- Conv2D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Dense
- DepthwiseConv2D
- Embedding
- LayerNormalization
- GRU
- LSTM
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- MutiHeadAttention
- SeparableConv1D
- SeparableConv2D
- MultiHeadAttention
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- SimpleRNN
- 1D
- 2D
- 3D
- 4D
- 5D
- 6D
- Scalar
- Show All Articles (22) Collapse Articles
-
- AdditiveAttention
- Attention
- BatchNormalization
- Conv1D
- Conv2D
- Conv1DTranspose
- Conv2DTranspose
- Bidirectional
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Conv3DTranspose
- DepthwiseConv2D
- Dense
- Embedding
- LayerNormalization
- GRU
- PReLU 2D
- PReLU 3D
- PReLU 4D
- MultiHeadAttention
- LSTM
- PReLU 5D
- SeparableConv1D
- SeparableConv2D
- SimpleRNN
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- 1D
- 2D
- 3D
- 4D
- 5D
- 6D
- Scalar
- Show All Articles (21) Collapse Articles
-
-
- AdditiveAttention
- Attention
- BatchNormalization
- Bidirectional
- Conv1D
- Conv2D
- Conv3D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Dense
- DepthwiseConv2D
- Embedding
- GRU
- LayerNormalization
- LSTM
- MultiHeadAttention
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- Resume
- SeparableConv1D
- SeparableConv2D
- SimpleRNN
- Show All Articles (12) Collapse Articles
-
-
- Dense
- Embedding
- AdditiveAttention
- Attention
- MultiHeadAttention
- Conv1D
- Conv2D
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- DepthwiseConv2D
- SeparableConv1D
- SeparableConv2D
- BatchNormalization
- LayerNormalization
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- Bidirectional
- GRU
- LSTM
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- SimpleRNN
- PReLU 4D
- Show All Articles (15) Collapse Articles
-
- Dense
- Embedding
- AdditiveAttention
- Attention
- MultiHeadAttention
- Conv1D
- Conv2D
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- DepthwiseConv2D
- SeparableConv1D
- SeparableConv2D
- BatchNormalization
- LayerNormalization
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- Bidirectional
- GRU
- LSTM
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- SimpleRNN
- Show All Articles (14) Collapse Articles
-
-
- Accuracy
- BinaryAccuracy
- BinaryCrossentropy
- BinaryIoU
- CategoricalAccuracy
- CategoricalCrossentropy
- CategoricalHinge
- CosineSimilarity
- FalseNegatives
- FalsePositives
- Hinge
- Huber
- IoU
- KLDivergence
- LogCoshError
- Mean
- MeanAbsoluteError
- MeanAbsolutePercentageError
- MeanIoU
- MeanRelativeError
- MeanSquaredError
- MeanSquaredLogarithmicError
- MeanTensor
- OneHotIoU
- OneHotMeanIoU
- Poisson
- Precision
- PrecisionAtRecall
- Recall
- RecallAtPrecision
- RootMeanSquaredError
- SensitivityAtSpecificity
- SparseCategoricalAccuracy
- SparseCategoricalCrossentropy
- SparseTopKCategoricalAccuracy
- Specificity
- SpecificityAtSensitivity
- SquaredHinge
- Sum
- TopKCategoricalAccuracy
- TrueNegatives
- TruePositives
- Resume
- Show All Articles (27) Collapse Articles
-
-
- Dense
- Embedding
- AdditiveAttention
- Attention
- MultiHeadAttention
- Conv1D
- Conv2D
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- DepthwiseConv2D
- SeparableConv1D
- SeparableConv2D
- BatchNormalization
- LayerNormalization
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- Bidirectional
- GRU
- LSTM
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- SimpleRNN
- Show All Articles (14) Collapse Articles
-
- Dense
- Embedding
- AdditiveAttention
- Attention
- MultiHeadAttention
- Conv1D
- Conv2D
- Conv3D
- ConvLSTM1D
- ConvLSTM2D
- ConvLSTM3D
- Conv1DTranspose
- Conv2DTranspose
- Conv3DTranspose
- DepthwiseConv2D
- SeparableConv1D
- SeparableConv2D
- LayerNormalization
- PReLU 2D
- PReLU 3D
- PReLU 4D
- PReLU 5D
- Bidirectional
- GRU
- LSTM
- RNN (GRU)
- RNN (LSTM)
- RNN (SimpleRNN)
- SimpleRNN
- BatchNormalization
- Show All Articles (14) Collapse Articles
-
-
-
Computer Vision Toolkit
-
CUDA Toolkit
-
- Resume
- Array size
- Index Array
- Replace Subset
- Insert Into Array
- Delete From Array
- Initialize Array
- Build Array
- Concatenate Array
- Array Subset
- Min & Max
- Reshape Array
- Short Array
- Reverse 1D array
- Shuffle array
- Search In Array
- Split 1D Array
- Split 2D Array
- Rotate 1D Array
- Increment Array Element
- Decrement Array Element
- Interpolate 1D Array
- Threshold 1D Array
- Interleave 1D Array
- Decimate 1D Array
- Transpose Array
- Remove Duplicate From 1D Array
- Show All Articles (11) Collapse Articles
-
-
- Resume
- Add
- Substract
- Multiply
- Divide
- Quotient & Remainder
- Increment
- Decrement
- Add Array Element
- Multiply Array Element
- Absolute
- Round To Nearest
- Round Toward -Infinity
- Round Toward +Infinity
- Scale By Power Of Two
- Square Root
- Square
- Negate
- Reciprocal
- Sign
- Show All Articles (4) Collapse Articles
DequantizeLinear
Description
The linear dequantization operator. It consumes a quantized tensor, a scale, and a zero point to compute the full-precision tensor. The dequantization formula is y = (x - x_zero_point) * x_scale. x_scale and x_zero_point must have the same shape, determining the quantization’s granularity: a scalar for per-tensor/per-layer quantization, a 1-D tensor for per-axis quantization, or have a rank identical to the input for blocked quantization. See QuantizeLinear for details on quantization granularity.
x_zero_point and x must have the same type. x and y must have the same shape. In the case of dequantizing int32, there’s no zero point (zero point is supposed to be 0). zero-point is usually not used in the case of float8 and 4-bit types quantization, but the dequantization formula remains the same for consistency. The output type is determined by the attribute output_dtype. If output_dtype is not supplied then the output type is the same as x_scale. The output type also determines the precision of the multiplication operation.
Input parameters
specified_outputs_name : array, this parameter lets you manually assign custom names to the output tensors of a node.
Graphs in : cluster, ONNX model architecture.
x (heterogeneous) – T1 : object, N-D quantized input tensor to be de-quantized. x_scale (heterogeneous) – T2 : object, scale for input x. For per-tensor/layer dequantization the scale is a scalar, for per per-axis dequantization it is a 1-D Tensor and for blocked dequantization it has the same shape as the input, except for one dimension in which blocking is performed. x_zero_point (optional, heterogeneous) – T1 : object, zero point for input x. Shape must match x_scale. It’s optional. Zero point is 0 when it’s not specified.
Parameters : cluster,
axis : integer, the axis of the dequantizing dimension of the input tensor. Used for per-axis and blocked quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).
Default value “0”.
training? : boolean, whether the layer is in training mode (can store data for backward).
Default value “True”.
lda coeff : float, defines the coefficient by which the loss derivative will be multiplied before being sent to the previous layer (since during the backward run we go backwards).
Default value “1”.
name (optional) : string, name of the node.
Output parameters
y (heterogeneous) – T3 : object, N-D full precision output tensor. It has the same shape as input x. The data type is specified by the output_dtype attribute or, in its absence, the type of x_scale.
Type Constraints
T1 in (tensor(float4e2m1), tensor(float8e4m3fn), tensor(float8e4m3fnuz), tensor(float8e5m2), tensor(float8e5m2fnuz), tensor(int16), tensor(int32), tensor(int4), tensor(int8), tensor(uint16), tensor(uint4), tensor(uint8)) : The type of the inputs ‘x_zero_point’ and ‘x’.
T2 in (tensor(bfloat16), tensor(float), tensor(float16), tensor(float8e8m0)) : The type of the input ‘x_scale’.
T3 in (tensor(bfloat16), tensor(float), tensor(float16)) : The type of the output ‘y’.