GPU Backend Op Definition Supplement

ArgbToRgb

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Argmax

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	out[0]
All	Shape: Rank in [1,4]	Shape: Rank > 0

Argmin

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	out[0]
All	Shape: Rank in [1,4]	Shape: Rank > 0

Batchnorm

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Rank in [1,4]

BatchToSpace

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	crops
All	Value: crops[0] and crops[1] should be equal. Value: crops[2] and crops[3] should be equal.

Cast

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_INT_64, QNN_DATATYPE_UINT_32, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_32, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

ChannelShuffle

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Shape: Rank in [3,4]

Support

Configuration
All	Param axis only supports default value

Concat

Datatypes

Configuration	in[0..m]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0..m]	axis
All	Shape: Rank in [1,5]	Value: If any in[0..m] dimension > 16384 and N = 4, then axis > 0 Value: If N = 5, then axis > 0 Value: If N = 5 and axis = 1, only batch = 1 supported

Conv2d

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Support

Configuration
All	Param reuse_sparse_indices only supports default value

CumulativeSum

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	axis
All	Value: Must be in range [0, rank(in[0]) - 1]. Datatype: QNN_DATATYPE_UINT_32

DepthToSpace

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	block_size
All	Value: block_size[0] and block_size[1] should be equal.

DepthWiseConv2d

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Dequantize

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_SFIXED_POINT_8, QNN_DATATYPE_UFIXED_POINT_8, QNN_DATATYPE_UFIXED_POINT_4	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Only supports ScaleOffset based quantization encoding type

DetectionOutput

Datatypes

Configuration	in[0]	in[1]	in[2]
All	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_FLOAT_32

Constraints

Configuration	in[0]	in[1]
All	Shape: Only batch = 1 supported Shape: num_anchors less than or equal to 2048 in Fast mode Shape: num_classes less than 256	Shape: num_boxes should be equal to num_anchors

Support

Configuration
All	Param share_location only supports default value Param nms_eta only supports default value

ElementWiseAbs

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

ElementWiseAdd

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Datatype: INT_32 supported only in broadcast Shape: Only supports Rank >= 1	Datatype: INT_32 supported only in broadcast Shape: Only supports Rank >= 1	Datatype: INT_32 supported only in broadcast Shape: Only supports Rank >= 1

ElementWiseAnd

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseBinary

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_8, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_8, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_8, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]	operation
All	Datatype: Numerical supports: FLOAT_32, FLOAT_16, INT_32. Comparison supports: FLOAT_32, FLOAT_16, INT_32, BOOL_8. Logical supports: UINT_8, BOOL_8 Shape: Only supports Rank >= 1 If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static	Datatype: Numerical supports: FLOAT_32, FLOAT_16, INT_32. Comparison supports: FLOAT_32, FLOAT_16, INT_32, BOOL_8. Logical supports: UINT_8, BOOL_8 Shape: Only supports Rank >= 1 If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static	Datatype: Numerical supports: FLOAT_32, FLOAT_16, INT_32. Comparison supports: UINT_8, BOOL_8. Logical supports: UINT_8, BOOL_8 Shape: Only supports Rank >= 1	Value: FMOD, MOD, and XOR not supported

ElementWiseCos

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseDivide

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseEqual

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseExp

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseFloor

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseFloorDiv

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

ElementWiseGreater

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseGreaterEqual

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseLess

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseLessEqual

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseLog

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseMaximum

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseMinimum

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseMultiply

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseNeg

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

ElementWiseNeuron

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseNot

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

ElementWiseNotEqual

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseOr

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8	QNN_DATATYPE_BOOL_8, QNN_DATATYPE_UINT_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWisePower

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseRound

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseRsqrt

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseSelect

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_UFIXED_POINT_4	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_UFIXED_POINT_4	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[1]	in[2]
All	If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static	If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static

ElementWiseSin

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseSoftplus

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseSquaredDifference

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseSquareRoot

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ElementWiseSubtract

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1	Shape: Only supports Rank >= 1

ElementWiseUnary

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	out[0]	operation
All	Datatype: BOOL_8 and UINT_8 data types only supported when operation param is NOT	Datatype: BOOL_8 and UINT_8 data types only supported when operation param is NOT	Value: ASIN, ATAN, RECIPROCAL, and SIGN not supported

Elu

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ExpandDims

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

FullyConnected

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	out[0]
All	Datatype: out[0] must be same Datatype as in[0]

Support

Configuration
All	Param keep_dims only supports default value

Gather

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	in[1]
All	If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static	Shape: k must be 1,2 or 3 If datatype is QNN_DATATYPE_UFIXED_POINT_4, then input must be static

GatherNd

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[1]	batch_dims
All	Shape: k must be 1,2 or 3 and Shape(in[1])[k-1] must be greater than 0 and less than or equal to (n - batch_dims) Value: Indices cannot be out of range of in[0]	Value: batch_dims must be less than min(n, k) Shape: Shape(in[0])[:batch_dims] must equal Shape(in[1])[:batch_dims]

Gelu

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

GridSample

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Input rank must be 4	Shape: Input rank must be 4 Datatype: Same Datatype as in[0]	Shape: Output rank must be 4 Datatype: Same Datatype as in[0]

GroupNorm

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Supports rank in [3,4]

HardSwish

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

HeatMapMaxKeyPoint

Datatypes

Configuration	in[0]	in[1]	out[0]	out[1]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Shape: Only square heat maps supported

InstanceNorm

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	mode	region
All	Supports rank in [3,4]	Value: Only supports mode MU_SIGMA	Value: Only supports region ACROSS_SPATIAL

Support

Configuration
All	Param normalize_variance only supports default value

L2Norm

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	axes
All	Unsupported

L2Pool2d

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

LayerNorm

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[1]	in[2]
All	Shape: Only supports input of Rank = 1 Shape: Dimension must be equal to norm axis dimension of in[0]	Shape: Only supports input of Rank = 1 Shape: Dimension must be equal to norm axis dimension of in[0]

LogSoftmax

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	axis
All	Value: Only support value of N-1

Lrn

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Lstm

Datatypes

Configuration	in[0]	in[1]	in[2]	in[3]	in[4]	in[5]	in[6]	in[7]	in[8]	in[9]	in[10]	in[11]	in[12]	in[13]	in[14]	in[15]	in[16]	in[17]	in[21]	in[22]	in[23]	in[24]	out[0]	out[1]	out[2]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	in[1]	in[2]	in[3]	in[4]	in[5]	in[6]	in[7]	in[8]	in[9]	in[12]	in[13]	in[14]	in[15]	in[16]	in[17]	in[18]	in[19]	in[20]	in[21]	in[22]	in[23]	in[24]	input_gate_qscale	forget_gate_qscale	cell_gate_qscale	output_gate_qscale	hidden_state_offset	hidden_state_qscale	time_major
All	Shape: Input rank must be 2.	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	input must be static	cell to input weights not supported	cell to forget weights not supported	cell to output weights not supported	input must be static	input must be static	input must be static	reset only supports default	Only support default value 0.0f	Only support default value 0.0f	Only support default value 0.0f	Only support default value 0.0f	Only support default value 0.0f	Only support default value 0.0f	Not applicable for 2D input and is ignored.

Support

Configuration
All	Param direction only supports default value

MatMul

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	in[1]	in[2]	out[0]	transpose_in0
All	Dynamic Shape: Dynamic dims not supported.	Shape: Batch dimensions (outermost N-2) if not equal to that of in[0], should = 1 or shouldn’t exist Dynamic Shape: Dynamic dims not supported.	Dynamic Shape: Dynamic dims not supported.	Shape: Rank must be equal to max(rank(in[0]), rank(in[1])) i.e after batch broadcasting Shape: Batch dimensions should be broadcasted from in[0], in[1] Dynamic Shape: Dynamic dims not supported.	Value: Must be false

Nv12ToRgb

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_FLOAT_32

Nv21ToRgb

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_FLOAT_32

Pack

Datatypes

Configuration	in[0..m]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0..m]	axis
All	Shape: N less than or equal to 3	Value: In range [0, rank(in[0])-1] for m == 1 Value: In range [max(0,rank(in[0])-2), rank(in[0])-1] for m >= 2

Pad

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	scheme	pad_constant_value
All	Value: MIRROR_SYMMETRIC can only be used when rank(in[0]) is 4 Value: MIRROR_SYMMETRIC can only be used with H or W dimensions Value: MIRROR_REFLECT can only be used when rank(in[0]) is 4 Value: EDGE can only be used when rank(in[0]) is less than or equal to 4	Datatype: QNN_DATATYPE_FLOAT_32

PoolAvg2d

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Support

Configuration
All	Param rounding_mode only supports default value

PoolMax2d

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Support

Configuration
All	Param rounding_mode only supports default value

Prelu

Datatypes

Configuration	in[0]	in[1]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	in[1]
All	Shape: N >= 1	Shape: M >= 1

ReduceMax

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	axes
All	Shape: N less than or equal to 4	Value: Reduction across batch supported only when rank(axes) == 1

ReduceMean

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	axes
All	Shape: N less than or equal to 4	Value: Reduction across batch supported only when rank(axes) == 1

ReduceMin

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	axes
All	Shape: N less than or equal to 4	Value: Reduction across batch supported only when rank(axes) == 1

ReduceProd

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	axes
All	Shape: N less than or equal to 4	Value: Reduction across batch supported only when rank(axes) == 1

ReduceSum

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	axes
All	Shape: N less than or equal to 4	Value: Reduction across batch supported only when rank(axes) == 1

Relu

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Relu1

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Relu6

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

ReluMinMax

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Reshape

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_8, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	in[1]	out[0]
All	Shape: Only supports Rank >= 1 Dynamic Shape: Dynamic dims not supported.	This input is not supported.	Shape: Only supports Rank >= 1 Dynamic Shape: Dynamic dims not supported.

Resize

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	out[0]
All	Shape: dim of out0[1] and out0[2] must be > 1

ResizeBilinear

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	out[0]
All	Shape: dim of out0[1] and out0[2] must be > 1

Support

Configuration
All	Param antialias only supports default value

ResizeNearestNeighbor

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	out[0]
All	Shape: dim of out0[1] and out0[2] must be > 1

RmsNorm

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]	in[1]	in[2]	axes
All	Shape: Only supports input of Rank in range [1, 4]	Shape: Only supports input of Rank = 1 Shape: Dimension must be equal to norm axis dimension of in[0]	Shape: Only supports input of Rank = 1 Shape: Dimension must be equal to norm axis dimension of in[0]	Value: Value should be in range [0, Rank-1]

RoiAlign

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_FLOAT_32	QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_32	QNN_DATATYPE_FLOAT_32

Support

Configuration
All	Param aligned only supports default value Param allow_invalid_roi only supports default value

ScatterElements

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_BOOL_8	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_BOOL_8

Constraints

Configuration	in[0]	axis
All	Shape: Rank in [1,4]	Value: If N=4, then axis > 0.

Sigmoid

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Softmax

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Support

Configuration
All	Param axis only supports default value

SpaceToBatch

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	pad_amount
All	Value: pad_amount[0] and pad_amount[1] should be equal. Value: pad_amount[2] and pad_amount[3] should be equal.

SpaceToDepth

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	block_size
All	Value: block_size[0] and block_size[1] should be equal.

Support

Configuration
All	Param mode only supports default value

Split

Datatypes

Configuration	in[0]	out[0]…out[m]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	axis
All	Value: Should be greater than N-4 for N > 4

Squeeze

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

StridedSlice

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_UINT_32, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]	ranges
All	Shape: Rank must be 5 or less	Value: No support for slicing across batch dimensions > 4, the slice ranges specified must align

Support

Configuration
All	Param new_axes_mask only supports default value

Tanh

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Tile

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	multiples
All	Value: For N > 3, multiples[0…N-4] should be 1

TopK

Datatypes

Configuration	in[0]	out[0]	out[1]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_INT_32, QNN_DATATYPE_UINT_32

Support

Configuration
All	Param largest only supports default value

Transpose

Datatypes

Configuration	in[0]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16, QNN_DATATYPE_INT_32

Constraints

Configuration	in[0]
All	Shape: Rank must be 2, 3, 4, or 5

TransposeConv2d

Datatypes

Configuration	in[0]	in[1]	in[2]	out[0]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Support

Configuration
All	Param group only supports default value

UnPack

Datatypes

Configuration	in[0]	out[0]…out[m]
All	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16	QNN_DATATYPE_FLOAT_32, QNN_DATATYPE_FLOAT_16

Constraints

Configuration	in[0]
All	Shape: Rank must be 2, 3, or 4