BNNSFilterCreateLayerEmbedding(_:_:)
Returns a new embedding layer.
Declaration
func BNNSFilterCreateLayerEmbedding(_ layer_params: UnsafePointer<BNNSLayerParametersEmbedding>, _ filter_params: UnsafePointer<BNNSFilterParameters>?) -> BNNSFilter?Parameters
- layer_params:
The layer parameters.
- filter_params:
The filter runtime parameters.
Discussion
Use this function in conjunction with BNNSLayerParametersEmbedding to create an embedding layer that accesses dictionary items based on the integer indices that the input tensor defines.
For example, the following code shows how to access three dictionary items at indices 1, 5, and 2. The dictionary contains seven items that are five-element vectors. Therefore, the indices descriptor has the shape (3), the dictionary descriptor has the shape (5, 7), and the output descriptor has the shape (5, 3).
static func embeddingLayer() {
let indicesData = UnsafeMutableBufferPointer<Int8>.allocate(capacity: 3)
_ = indicesData.initialize(from: [1, 5, 2])
let indicesDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayoutVector,
size: (3, 0, 0, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: indicesData.baseAddress!,
data_type: BNNSDataType.int8,
table_data: nil,
table_data_type: BNNSDataType.int8,
data_scale: 1, data_bias: 0)
let dictionaryItemSize = 5
let dictionaryItemCount = 7
let dictionaryData = UnsafeMutableBufferPointer<Float>.allocate(capacity: dictionaryItemSize * dictionaryItemCount)
_ = dictionaryData.initialize(from: [01, 02, 03, 04, 05,
11, 12, 13, 14, 15,
21, 22, 23, 24, 25,
31, 32, 33, 34, 35,
41, 42, 43, 44, 45,
51, 52, 53, 54, 55,
61, 62, 63, 64, 65])
let dictionaryDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayout2DLastMajor,
size: (dictionaryItemSize, dictionaryItemCount,
0, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: dictionaryData.baseAddress!,
data_type: BNNSDataType.float,
table_data: nil,
table_data_type: BNNSDataType.float,
data_scale: 1, data_bias: 0)
let outputData = UnsafeMutableBufferPointer<Float>.allocate(capacity: dictionaryItemSize * indicesData.count)
let outputDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayout2DLastMajor,
size: (dictionaryItemSize, indicesData.count,
0, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: outputData.baseAddress!,
data_type: BNNSDataType.float,
table_data: nil,
table_data_type: BNNSDataType.float,
data_scale: 1, data_bias: 0)
var parameters = BNNSLayerParametersEmbedding(flags: BNNSEmbeddingFlags(rawValue: 0),
i_desc: indicesDescriptor,
o_desc: outputDescriptor,
dictionary: dictionaryDescriptor,
padding_idx: 0,
max_norm: 0,
norm_type: 0)
let embeddingLayer = BNNSFilterCreateLayerEmbedding(¶meters, nil)
BNNSFilterApply(embeddingLayer,
indicesDescriptor.data!,
outputDescriptor.data!)
// Prints:
// [11.0, 12.0, 13.0, 14.0, 15.0,
// 51.0, 52.0, 53.0, 54.0, 55.0,
// 21.0, 22.0, 23.0, 24.0, 25.0]
print(Array(outputData))
indicesData.deallocate()
dictionaryData.deallocate()
outputData.deallocate()
}The embedding layer supports clipping to a maximum norm. The following code accesses the second and third items from a dictionary that contains three 2 x 3 matrices. The code initializes the embedding layer with a maximum norm that is the infinity-norm of the first dictionary item.
static func embeddingLayerClipByNorm() {
let indicesData = UnsafeMutableBufferPointer<Int8>.allocate(capacity: 2)
_ = indicesData.initialize(from: [1, 2])
let indicesDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayoutVector,
size: (2, 0, 0, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: indicesData.baseAddress!,
data_type: BNNSDataType.int8,
table_data: nil,
table_data_type: BNNSDataType.float,
data_scale: 1, data_bias: 0)
let dictionaryItemRowCount = 2
let dictionaryItemColumnCount = 3
let dictionaryItemCount = 3
let dictionaryElementCount = dictionaryItemRowCount * dictionaryItemColumnCount * dictionaryItemCount
let dictionaryData = UnsafeMutableBufferPointer<Float>.allocate(capacity: dictionaryElementCount)
_ = dictionaryData.initialize(from: [0.1, 0.2, 0.3,
0.4, 0.5, 0.6,
1, 3, 5,
2, 4, 6,
60, 50, 40,
30, 20, 10])
let dictionaryDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayout3DLastMajor,
size: (dictionaryItemRowCount,
dictionaryItemColumnCount,
dictionaryItemCount, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: dictionaryData.baseAddress!,
data_type: BNNSDataType.float,
table_data: nil,
table_data_type: BNNSDataType.float,
data_scale: 1, data_bias: 0)
let outputElementCount = dictionaryItemRowCount * dictionaryItemColumnCount * indicesData.count
let outputData = UnsafeMutableBufferPointer<Float>.allocate(capacity: outputElementCount)
let outputDescriptor = BNNSNDArrayDescriptor(flags: BNNSNDArrayFlags(0),
layout: BNNSDataLayout3DLastMajor,
size: (dictionaryItemRowCount,
dictionaryItemColumnCount,
indicesData.count, 0, 0, 0, 0, 0),
stride: (0, 0, 0, 0, 0, 0, 0, 0),
data: outputData.baseAddress!,
data_type: BNNSDataType.float,
table_data: nil,
table_data_type: BNNSDataType.float,
data_scale: 1, data_bias: 0)
var parameters = BNNSLayerParametersEmbedding(flags: BNNSEmbeddingFlags(rawValue: 0),
i_desc: indicesDescriptor,
o_desc: outputDescriptor,
dictionary: dictionaryDescriptor,
padding_idx: 0,
max_norm: 0.6,
norm_type: .infinity)
let embeddingLayer = BNNSFilterCreateLayerEmbedding(¶meters, nil)
BNNSFilterApply(embeddingLayer,
indicesDescriptor.data!,
outputDescriptor.data!)
// Prints:
// [0.1, 0.3, 0.5,
// 0.2, 0.4, 0.6,
//
// 0.6, 0.5, 0.4,
// 0.3, 0.2, 0.1]
print(Array(outputData))
indicesData.deallocate()
dictionaryData.deallocate()
outputData.deallocate()
}