• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# The layers that ArmNN SDK Serializer currently supports.
2
3This reference guide provides a list of layers which can be serialized currently by the Arm NN SDK.
4
5## Fully supported
6
7The Arm NN SDK Serializer currently supports the following layers:
8
9* Activation
10* Addition
11* ArgMinMax
12* BatchToSpaceNd
13* BatchNormalization
14* Comparison
15* Concat
16* Constant
17* Convolution2d
18* DepthToSpace
19* DepthwiseConvolution2d
20* Dequantize
21* DetectionPostProcess
22* Division
23* ElementwiseUnary
24* Fill
25* Floor
26* FullyConnected
27* Gather
28* Input
29* InstanceNormalization
30* L2Normalization
31* Logical
32* LogSoftmax
33* Lstm
34* Maximum
35* Mean
36* Merge
37* Minimum
38* Multiplication
39* Normalization
40* Output
41* Pad
42* Permute
43* Pooling2d
44* Prelu
45* QLstm
46* Quantize
47* QuantizedLstm
48* Rank
49* Reshape
50* Resize
51* Slice
52* Softmax
53* SpaceToBatchNd
54* SpaceToDepth
55* Splitter
56* Stack
57* StandIn
58* StridedSlice
59* Subtraction
60* Switch
61* Transpose
62* TransposeConvolution2d
63
64More machine learning layers will be supported in future releases.
65
66## Deprecated layers
67
68Some layers have been deprecated and replaced by others layers. In order to maintain backward compatibility, serializations of these deprecated layers will deserialize to the layers that have replaced them, as follows:
69
70* Abs will deserialize as ElementwiseUnary
71* Equal will deserialize as Comparison
72* Greater will deserialize as Comparison
73* Merger will deserialize as Concat
74* ResizeBilinear will deserialize as Resize
75* Rsqrt will deserialize as ElementwiseUnary
76
77