S100/S100P ONNX算子BPU约束列表
下方表格中：
lhs：left-hand side，指运算中的左操作数。
rhs：right-hand side，指运算中的右操作数。
注解
- DeformConv算子本身存在一定特殊性，
我们通过特定的自定义算子的方法实现了对
torchvision::deformconv的直接支持， 如有需要请联系地平线技术支持人员获取相关说明。 - GroupNormalization算子在从Pytorch框架导出到ONNX时， 会被框架自动拆分成多个子算子到原始ONNX模型中， 我们支持对这些子算子的BPU计算，从而实现了对GroupNormalization算子的支持。
| ONNX Operator Name | BPU Support Constraints |
|---|---|
| Abs | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] | |
| Acos | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Acosh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Add | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| And | lhs: Type: int8, int16, bool8, float16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| ArgMax | input: Type: int8, int16, int32, float16, float32 Shape: [*] Dim: if type is int8, int16 or float16, reduce axis dim size ∈ [1, 32767] Element : if type is int8 or int16, reduce Elements size ∈ [1, 65535] dims: If type is int32, float16, float32, only support one dim. output: Type: int8, int16, int32 Shape: reduce dim will be 1 or fused depend on keepDims |
| ArgMin | input: Type: int8, int16, int32, float16, float32 Shape: [*] Dim: if type is int8, int16 or float16, reduce axis dim size ∈ [1, 32767] Element : if type is int8 or int16, reduce Elements size ∈ [1, 65535] dims: If type is int32, float16, float32, only support one dim. output: Type: int8, int16, int32 Shape: reduce dim will be 1 or fused depend on keepDims |
| Asin | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Asinh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Atan | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Atanh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| AveragePool | input: Type: int8, int16 Shape: [*,H,W,C] or [*,L,C] output: The input and output types need to be the same. kernel: Shape: [KL] or [KH,KW], only support 1d or 2d now Dim: 1d: KL ∈ [1, 256], KL*bitWidth/8 <= 24576; 2d: KH, KW ∈ [1, 256], KH*KW*bitWidth/8 <= 24576 stride: Shape: [SH,SW] or [SL] Dim: SH, SW, SL ∈ [1, 256] pad: Shape: [PH_BEGIN,PW_BEGIN,PH_END,PW_END] or [PL_BEGIN,PL_END] PH_BEGIN,PW_BEGIN,PL_BEGIN,PH_END,PW_END,PL_END ∈ [-255, 256] dilation: Shape: 1d: [DW]; 2d: [DH, DW] Dim: 1d: DW ∈ {1}; 2d: DH, DW ∈ {1} |
| BatchNormalization | N/A, collapsed in graph optimization phase |
| BitwiseAnd | lhs: Type: int8, int16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| BitwiseNot | input: Type: int8, int16 Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| BitwiseOr | lhs: Type: int8, int16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| BitwiseXor | lhs: Type: int8, int16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Cast | input: Type: int8, int16, int32, float16, float32, bool8 Shape: [*] output: Type: int8, int16, int32, float16, float32, bool8. Not support combinations (float16,float32→bool8). Shape: [*] |
| Ceil | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| otherwise: input: Type: int8, int16, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] | |
| Celu | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Clip | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] | |
| Concat | input: Type: No limits Arg Number: input number ∈ [1, 1024] Dim: all dims < 131072 Size: size < 2G output: The input and output types need to be the same. Dim: all dims < 131072 Size: size < 2G |
| Constant | N/A, collapsed in graph optimization phase |
| ConstantOfShape | N/A, collapsed in graph optimization phase |
| Conv | input: --conv 1d-- Type: int8, int16 Shape: [*,L,C] Dim: * ∈ [1, 4096]; L,C ∈ [1, 65536] --conv 2d-- Type: int8, int16 Shape: [*,H,W,C] Dim: * ∈ [1, 4096]; H,W,C ∈ [1, 65536] --conv 3d-- Type: int8 Shape: [*,D,H,W,C] Dim: * ∈ [1, 128]; D,H,W ∈ [1, 65536]; C ∈ [1, 4096]; weight: --conv 1d-- Type: int8, int16 Shape: [N,KL,C] Dim: C ∈ [1, 8192]; KL ∈ [1, 31]; N ∈ [1, 65536] if fout is the last layer of conv else [1, 8192] Size: KL × C ∈ [1, 65535] --conv 2d-- Type: int8, int16 Shape: [N,KH,KW,C] Dim: C ∈ [1, 8192]; KH,KW ∈ [1, 31]; N ∈ [1, 65536] if fout is the last layer of conv else [1, 8192] Size: KH × KW × C ∈ [1, 65535] --conv 3d-- Type: int8 Shape: [N,KD,KH,KW,C] N ∈ [1, 65535]; KD,KH,KW ∈ [1, 9]; Dim: C ∈ [1, 4096]; Size: KD × KH × KW × C ∈ [1, 131072] bias: Type: float32 output: --conv 1d-- Type: int8, int16, int32 Shape: [*,L,C] Dim: * ∈ [1, 4096]; L,C ∈ [1, 65536] --conv 2d-- Type: int8, int16, int32 Shape: [*,H,W,C] Dim: * ∈ [1, 4096]; H,W,C ∈ [1, 65536] --conv 3d-- Type: int8, int16, int32 Shape: [*,D,H,W,C] Dim: * ∈ [1, 128]; D,H,W ∈ [1, 65536]; C ∈ [1, 4096]; stride: --conv 1d-- Shape: [SL] Dim: SL ∈ [1, 256]; SL ∈ {1} if dilation > 1 --conv 2d-- Shape: [SH,SW] Dim: SH,SW ∈ [1, 256]; SH,SW ∈ {1} if dilation > 1 --conv 3d-- Shape: [SD,SH,SW] Dim: SD,SH,SW must be 1 or 2 and equal to each other. pad: --conv 1d-- Shape: [P_left,P_right] Dim: P_left,P_right ∈ (-L, 256] --conv 2d-- Shape: [P_top,P_left,P_bottom,P_right] Dim: P_top,P_bottom ∈ (-H, 256], P_left,P_right ∈ (-W, 256] --conv 3d-- Shape: [P_front, P_top, P_left, P_back, P_bottom, P_right] Dim: P_front,P_back ∈ [0, KD/2], P_top,P_bottom ∈ [0, KH/2], P_left,P_right ∈ [0, KW/2] groupNum: Fin.c is divisible by group number, conv 3d only support 1 dilation: --conv 1d-- Shape: [DL] Dim: DL ∈ [1, 18] --conv 2d-- Shape: [DH,DW] Dim: DH,DW ∈ [1, 18] --conv 3d-- Shape: [DD,DH,DW] DD,DH,DW = 1 others: --conv 1d-- Stride only support odd number and 2 when conv is a int16 depthwise conv If groupNum > 1, for each group, fin.c' ∈ [1, 65535], KL × fin.c' ∈ [1, 65535] --conv 2d-- Stride only support odd number and 2 when conv is a int16 depthwise conv If groupNum > 1, for each group, fin.c' ∈ [1, 65535], KH × KW × fin.c' ∈ [1, 65535] Fin.c' = fin.c × min(lcm(fout.c × (lcm(fin.c, 4) / fin.c), 8) / fout.c, groupNum) |
| ConvTranspose | input: --convtranspose 1d/2d-- Type: int8, int16; input and weight cannot both be int16 1d_Shape: [*,W,C] 1d_Dim: * ∈ [1, 128]; W ∈ [1, 65536]; C ∈ [1, 2048] 2d_Shape: [*,H,W,C] 2d_Dim: * ∈ [1, 128]; H,W ∈ [1, 65536]; C ∈ [1, 2048] --convtranspose 3d-- Type: int8 3d_Shape: [*,D,H,W,C] 3d_Dim: * ∈ [1, 128]; D,H,W ∈ [1, 65536]; C ∈ [1, 2048] weight: --convtranspose 1d/2d-- Type: int8, int16; input and weight cannot both be int16 1d_Shape: [N,KW,C] 1d_Dim: N,C ∈ [1, 2048]; KW ∈ [1, 14] 1d_Size: KW × C ∈ [1, 65535] 2d_Shape: [N,KH,KW,C] 2d_Dim: N,C ∈ [1, 2048]; KH,KW ∈ [1, 14]; KH,KW cannot both be 1 2d_Size: KH × KW × C ∈ [1, 65535] --convtranspose 3d-- Type: int8 3d_Shape: [N,KD,KH,KW,C] 3d_Dim: N,C ∈ [1, 2048]; KD,KH,KW ∈ [1, 14]; KD,KH,KW cannot all be 1 3d_Size: KD × KH × KW × C ∈ [1, 131072] bias: Type: float32 output: Type: int8, int16, int32 stride: 1d_Shape: [SW] 1d_Dim: SW ∈ [1, 14]; 2d_Shape: [SH,SW] 2d_Dim: SH,SW ∈ [1, 14]; 3d_Shape: [SD,SH,SW] 3d_Dim: SD,SH,SW ∈ [1, 14]; pad: 1d_Shape: [P_left,P_bottom] 1d_Dim: P_left,P_bottom ∈ [0, 256] 2d_Shape: [P_top,P_left,P_bottom,P_right] 2d_Dim: P_top,P_left,P_bottom,P_right ∈ [0, 256] 3d_Shape: [P_front,P_top,P_left,P_back,P_bottom,P_right] 3d_Dim: P_front,P_top,P_left,P_back,P_bottom,P_right ∈ [0, 256] groupNum: Fin.c is divisible by group number, conv 3d only support 1 dilation: 1d_Shape: [DW] 1d_Dim: DW ∈ {1} 2d_Shape: [DH,DW] 2d_Dim: DH,DW ∈ {1} 3d_Shape: [DD,DH,DW] 3d_Dim: DD,DH,DW ∈ {1} |
| Cos | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Cosh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| CumSum | input: Type: int8, int16, float16, float32 Shape: [*, dim[axis], *] Dim: * ∈ [1, 65536]; dim[axis] ∈ [1, 8192] output: Type: int8, int16, int32, float16, float32. only support combinations (int8→int8,int16,int32), (int16→int16,int32), (float16→float16), (float32→float32) The Shape and Dim is same as input exclusive: If type is float16, float32, only support exclusive == 0, otherwise, exclusive is 0 or 1 reverse: If type is float16, float32, only support reverse == 0, otherwise, exclusive is 0 or 1 |
| DeformConv | input: Type: int8 Shape: [*,H,W,C] Dim: H,W ∈ [1, 1024]; H × W ≤ 720 × 1024; other dims ∈ [1, 65536] offset: Type: int16 Shape: [*,OH,OW,2 × offsetGroupNum × KH × KW] Size: 2 × offsetGroupNum × KH × KW ∈ [2, 256], OH × KH × OW × KW ≤ 720 × 1024 mask: Type: int8 Shape: [*,OH,OW,offsetGroupNum × KH × KW] Size: offsetGroupNum × KH × KW ∈ [1, 128] The value of mask is usually [0, 1] weight: Type: int8 Shape: [N,KH,KW,C] Dim: C ∈ [1, 8192]; KH,KW ∈ [1, 8]; N ∈ [1, 4096] Size: KH × KW × C ∈ [1, 65535] bias: Type: float32 output: Type: int8, int16, int32 Other constraints: Same as fin stride: Shape: [SH,SW] Dim: SH,SW ∈ [1] pad: Shape: [P_top,P_left,P_bottom,P_right] Dim: P_top,P_bottom ∈ [-H/2, 256], P_left,P_right ∈ [-W/2, 256] groupNum: Fin.c is divisible by group number offsetGroupNum: Fin.c is divisible by offset group number Size: offsetGroupNum ∈ [1, 2] dilation: Shape: [DH,DW] Dim: DH,DW ∈ [1] others: For each group, fin.c ∈ [1, 8192], KH × KW × fin.c ∈ [1, 65535], fin.c = C when group = 1 |
| DepthToSpace | input: Type: No limits output: The input and output types need to be the same. |
| Div | lhs: Type: quantized type support int8, int16, int32, others support int16, int32, float16, float32 rhs: The lhs and rhs types need to be the same. Shape: [*] output: For quantized type, support (int8,int16,int32→int8,int16,int32), others the input and output types need to be the same. Shape: [*] rounding_mode: For integer, only support TRUNC, For float, only support NONE. others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| Dropout | N/A, collapsed in graph optimization phase |
| Elu | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Equal | lhs: Type: int8, int16, int32, float16, float32, bool8 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| Erf | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Exp | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Expand | input: Type: No limits output: The input and output types need to be the same. |
| EyeLike | input: Type: int8, int16, int32, float16, float32, bool8 Shape: [*] output: Type: int8, int16, int32, float16, float32, bool8. Not support combinations (float16,float32→bool8). Shape: [*] |
| Floor | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| otherwise: input: Type: int8, int16, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] | |
| GRU | input: Type: int8, int16 Dim: all dims < 2097152 size < 2G output: Type: The input and output types need to be the same. |
| Gather | input: Type: int8, int16, int32, float16, float32 Shape: [*] Input will transpose to [N, W, C]. W is inputShape[dim], N is the product of inputShape[:dim], C is the product of inputShape[dim+1:]. N, C ∈ [1, 1048576], W ∈ [1, 4096]. If input type is int8, int16, W ∈ [1, 32768]. index: Type: int8, int16, int32; Unsupported negative indices. Shape: [*] index value should not be larger than 32768. And the reduce multiple of all index dims of shape should in range [1, 737280(720*1024)], because all dims Will be reduced to W dim of indices and output. If W of fout is larger than 737280, this op will be split too many sub-ops. output: The input and output types need to be the same. |
| GatherElements | input: Type: int8, int16, int32, float16, float32 Shape: [*] Input will transpose to [N, W, C]. W is inputShape[dim], N is the product of inputShape[:dim], C is the product of inputShape[dim+1:]. N, C ∈ [1, 1048576]. N × C should not be larger than 1048576 W ∈ [1, 4096]. If input type is int8, int16, W ∈ [1, 32768]. indices: Type: int8, int16, int32; Unsupported negative indices. Shape: [*] indices value should not be larger than 32768 Indices will transpose to [N, D, C]. D is indicesShape[dim], N is the product of indicesShape[:dim], C is the product of indicesShape[dim+1:]. N, C ∈ [1, 1048576], D ∈ [1, 737280(720*1024)]. IndicesShape[i] <= inputShape[i] for all dimensions i != dim. output: The input and output types need to be the same. |
| GatherND | input: Type: int8, int16, int32, float16, float32 Shape: [*] If gather1d, W = inputShape[batchDim], W ∈ [1, 4096]. If input type is int8, int16 W ∈ [1, 32768] If gather2d, H = inputShape[batchDim], W = inputShape[batchDim+1], H, W ∈ [1, 4096]. If input type is int8, int16, H, W ∈ [1, 32768]. H and W cannot both be greater than 4096 at the same time. B is product of inputShape[0: batchDim], B ∈ [1, 1048576]. C is product of inputShape[batchDim+D:], C ∈ [1, 1048576]. indices: Type: int8, int16, int32; Unsupported negative indices. Shape: [*, D] indices value should not be larger than 32768. D ∈ [1, 2]. Size: I is product of indicesShape[batchDim: batchDim+D], I ∈ [1, 737280]. output: Shape: [*] The input and output types need to be the same. batchDim: The number of batch dimensions. The gather of indexing starts from dimension of input[batchDim:] |
| Gelu | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Gemm | lhs: Type: int8, int16 Shape: [M,K] Dim: M,K ∈ [1, 8192] rhl: Type: int8, int16 Shape: [K,N] Dim: K ∈ [1, 8192], N ∈ [1, 1048576] output: Type: int8, int16, int32 Shape: [M,N] Other constraints: Same as lhs and rhs |
| GlobalAveragePool | input: Type: int8, int16 float16, float32 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: int8, int16, float16, float32. only support combinations (int8→int8,int16), (int16→int16), (float16→float16), (float32→float32) Shape: reduce dim will be 1 or fused depend on keepDim dims: If type is float16 or float32, only support one dim. |
| GlobalMaxPool | input: Type: int8, int16, int32, float16, float32 Shape: [*] Dim: if type is int8 or int16, reduce axis dim size ∈ [1, 65535] Element : if type is int8 or int16, reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. Shape: reduce dim will be 1 or fused depend on keepDim |
| Greater | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| GreaterOrEqual | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| GridSample | input: Type: int8 Shape: [*,H,W,C] Dim: H ∈ [1, 32768], W ∈ [1, 32768], other dims ∈ [1, 65536]. NOTE: H and W cannot both be greater than 4096 at the same time. grid: Type: int16 Shape: [*,H,W,2] output: Same as input except Dim constraints mode: Only support bilinear and nearest padding_mode: Only support zeros and border |
| GroupNormalization | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| HardSigmoid | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| HardSwish | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Identity | N/A, collapsed in graph optimization phase |
| InstanceNormalization | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| LSTM | input: Type: int8, int16 Dim: all dims < 2097152 size < 2G output: Type: The input and output types need to be the same. |
| LayerNormalization | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| LeakyRelu | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16, float16 Shape: [*] output: Type: int8, int16, float16. only support combinations (int8→int8), (int16→int8,int16), (float16->float16) Shape: [*] | |
| Less | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| LessOrEqual | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| Log | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| LogSoftmax | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| MatMul | lhs: Type: int8, int16 Shape: [*,M,C] Dim: * ∈ [1, 4096], M,C ∈ [1, 8192] rhl: Type: int8, int16 Shape: [*,C,N] Dim: * ∈ [1, 4096]; C ∈ [1, 8192], N ∈ [1, 1048576] output: Type: int8, int16, int32 Shape: [*,M,N] Other constraints: Same as lhs and rhs |
| Max | lhs: Type: int8, int16, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| MaxPool | input: Type: int8, int16 Shape: [*,H,W,C] output: The input and output types need to be the same. kernel: Shape: [KL] or [KH,KW], only support 1d or 2d now Dim: 1d: KL ∈ [1, 256], KL*bitWidth/8 <= 24576; 2d: KH, KW ∈ [1, 256], KH*KW*bitWidth/8 <= 24576 stride: Shape: [SH,SW] or [SL] Dim: SH, SW, SL ∈ [1, 256] pad: Shape: [PH_BEGIN,PW_BEGIN,PH_END,PW_END] or [PL_BEGIN,PL_END] PH_BEGIN,PW_BEGIN,PL_BEGIN,PH_END,PW_END,PL_END ∈ [-255, 256] dilation: Shape: 1d: [DW]; 2d: [DH, DW] Dim: 1d: DW ∈ {1}; 2d: DH, DW ∈ {1} |
| Min | lhs: Type: int8, int16, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Mish | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Mod | lhs: Type: int16, int32. Not support quantized type. Shape: [*] Value range: must be none-negative. rhs: The lhs and rhs types need to be the same. Shape: [*] Value range: must be positive. output: The input and output types need to be the same. Shape: [*] |
| Mul | lhs: Type: int8, int16, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: int8, int16, int32, float16, float32. only support (int8,int16→int8,int16,int32), (float16→float16), (float32→float32). Shape: [*] others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| Neg | input: Type: int8, int16, int32, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Not | input: Type: int8, int16, bool8, float16 Shape: [*] output: Type: bool8 Shape: [*] |
| Or | lhs: Type: int8, int16, bool8, float16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |
| PRelu | lhs: Type: int8, int16 Shape: [*] rhs: The type and shape is same as lhs output: The type and shape is same as lhs |
| Pad | N/A, collapsed in graph optimization phase |
| Pow | if exponent is an initializer with a single element 2, and the qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if exponent is an initializer with a single element 2, and the qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Reciprocal | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16, float16, float32 Shape: [*] output: Type: int8, int16, float16, float32. only support combinations (int8→int8), (int16→int8,int16), (float16->float16), (float32->float32) Shape: [*] | |
| ReduceL2 | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| ReduceMax | input: Type: int8, int16, int32, float16, float32 Shape: [*] Dim: if type is int8 or int16, reduce axis dim size ∈ [1, 65535] Element : if type is int8 or int16, reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. Shape: reduce dim will be 1 or fused depend on keepDim |
| ReduceMean | input: Type: int8, int16 float16, float32 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: int8, int16, float16, float32. only support combinations (int8→int8,int16), (int16→int16), (float16→float16), (float32→float32) Shape: reduce dim will be 1 or fused depend on keepDim dims: If type is float16 or float32, only support one dim. |
| ReduceMin | input: Type: int8, int16, int32, float16, float32 Shape: [*] Dim: if type is int8 or int16, reduce axis dim size ∈ [1, 65535] Element : if type is int8 or int16, reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. Shape: reduce dim will be 1 or fused depend on keepDim |
| ReduceSum | input: Type: int8, int16 float16, float32 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: int8, int16, float16, float32. only support combinations (int8→int8,int16), (int16→int16), (float16→float16), (float32→float32) Shape: reduce dim will be 1 or fused depend on keepDim dims: If type is float16 or float32, only support one dim. |
| Relu | input: Type: int8, int16, int32, float16 if type is int32, this op must be fusible to a Conv op Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Reshape | input: Type: No limits output: The input and output types need to be the same. |
| Resize | input: --case 1: int8-- Type: int8 Shape: [*,H,W,C] --case 2: float16, int16-- Type: float16, int16(with per tensor quant info) Shape: [*,H,W,C] Dim: H ∈ [1, 32768], W ∈ [1, 32768], C ∈ [1, 65536], other dims ∈ [1, 16384]. H and W cannot both be greater than 4096 at the same time. KN*KH*KW*KC*bitWidth/8 <= 786432 output: The input and output types need to be the same. mode: If Type is int8, support nearest and bilinear If Type is float16 or int16, only support bilinear padValue: If Type is int8, when padvalue is not equal to 0 and quantized, input only supports per tensor quantization step: If Type is int8, the integer part of step ∈ [-256, 255] If Type is float16 or int16, step ∈ [-8, 8] expansionMode: If Type is int8, support border and constant If Type is float16 or int16, only support border |
| Round | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Selu | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Shape | N/A, collapsed in graph optimization phase |
| Sigmoid | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Sign | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: float16 Shape: [*] output: The shape and type is same as input | |
| Sin | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Sinh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Size | N/A, collapsed in graph optimization phase |
| Slice | if starts/ends/axes/steps are constant: input: Type: No limits Dim: all dims < 2097152 output: Type: The output datatype must be the same as the input Dim: all dims < 2097152 |
| Softmax | input: Type: int8, int16 Shape: [*] Dim: reduce axis dim size ∈ [1, 65535] Element : reduce Elements size ∈ [1, 65535] output: Type: The input and output types need to be the same. |
| Softplus | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16 Shape: [*] output: Type: int8, int16. only support combinations (int8→int8), (int16→int8,int16) Shape: [*] | |
| SpaceToDepth | input: Type: No limits output: The input and output types need to be the same. |
| Split | input: Type: No limits Dim: all dims < 2097152 output: The input and output types need to be the same. The input and output constraints are the same. |
| Sqrt | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| otherwise: input: Type: int8, int16, float16, float32 Shape: [*] output: Type: int8, int16, float16, float32. only support combinations (int8→int8), (int16→int8,int16), (float16->float16), (float32->float32) Shape: [*] | |
| Sub | lhs: Type: int8, int16, float16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| Sum | lhs: Type: int8, int16, int32, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| Tan | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| Tanh | if qtype is int8: input: Type: int8, int16 table: The table and input types need to be the same. output: The input and output types need to be the same. |
| if qtype is int16: input: Type: int16 range: Type: int16 Shape: [N + 2, 4] N: * ∈ [1, 64] entry: Type: int16 Shape: [N, 64] N: * ∈ [1, 64] output: Type: int16 | |
| ThresholdedRelu | input: Type: int8, int16, float16 Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Tile | input: Type: No limits output: The input and output types need to be the same. |
| TopK | input: Type: int8, int16, int32, float16, float32 output: The type of output value is same as input, and indices is integer type. k: K <= 1024 others: The combined size in bytes of the op's operands and outputs must be no more than 4.8MB. |
| Transpose | input: Type: No limits output: The input and output types need to be the same. |
| Trilu | lhs: Type: int8, int16, float16, float32 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: int8, int16, int32, float16, float32. only support (int8,int16→int8,int16,int32), (float16→float16), (float32→float32). Shape: [*] others: When inputs are of integer or quantized type, if the result overflows, it will saturate to the maximum or minimum value of the result data type. |
| Where | condition: Type: bool8 Shape: [*] lhs: Type: int8, int16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: The input and output types need to be the same. Shape: [*] |
| Xor | lhs: Type: int8, int16, bool8, float16 Shape: [*] rhs: The lhs and rhs types need to be the same. Shape: [*] output: Type: bool8 Shape: [*] |