nuget
ITensorCompute 的标量兜底实现。
TensorComputeBase
01 Syntax
Microsoft.VisualBasic.MachineLearning.TensorFlow.Compute.TensorComputeBase
02 Methods
| Name | Overloads | Summary |
|---|---|---|
| RequireSameShape | 1 | 要求两个张量形状一致 |
| Wrap | 1 | 由计算结果的一维数组与形状包装出新的张量 |
| MapUnary | 1 | 逐元素一元运算的通用标量实现 |
| MapBinary | 1 | 逐元素二元运算的通用标量实现 |
| Heaviside | 1 | 阶跃函数(Heaviside): 大于 0 的元素取 1, 否则取 0 |
| ConvOutSize | 1 | 卷积/池化的输出边长: (input + 2*padding - kernel) / stride + 1 |
| RequireRank4 | 1 | 卷积/池化的入参形状校验 |
| ReduceAlongAxis | 1 | 沿指定轴执行规约操作（通用实现） |
| ReduceArgAxis | 1 | 沿指定轴找到极值的索引（argmax / argmin） |
| Add | 1 | |
| Subtract | 1 | |
| Multiply | 1 | |
| Divide | 1 | |
| Maximum | 1 | |
| Minimum | 1 | |
| Exp | 1 | |
| Log | 1 | |
| Sqrt | 1 | |
| Square | 1 | |
| Abs | 1 | |
| Sin | 1 | |
| Cos | 1 | |
| Tanh | 1 | |
| Sigmoid | 1 | |
| Negate | 1 | |
| Reciprocal | 1 | |
| Relu | 1 | |
| LeakyRelu | 1 | |
| Elu | 1 | |
| Gelu | 1 | |
| Swish | 1 | |
| AddScalar | 1 | |
| MultiplyScalar | 1 | |
| DivideScalar | 1 | |
| Pow | 1 | |
| Clip | 1 | |
| MatMul | 1 | |
| Transpose | 1 | |
| Conv2D | 1 | |
| Conv2DBackwardInput | 1 | |
| Conv2DBackwardFilter | 1 | |
| Conv2DBackwardBias | 1 | |
| MaxPool2D | 1 | |
| MaxPool2DBackward | 1 | |
| SumAll | 1 | |
| MeanAll | 1 | |
| Sum | 1 | |
| Mean | 1 | |
| Max | 1 | |
| Min | 1 | |
| Prod | 1 | |
| StdDev | 1 | |
| L2Norm | 1 | |
| ArgMax | 1 | |
| ArgMin | 1 | |
| Softmax | 1 | |
| LogSoftmax | 1 | |
| SigmoidCrossEntropyWithLogits | 1 | |
| MseLoss | 1 | |
| L2Loss | 1 | |
| HuberLoss | 1 |
03 Properties
| Name | Overloads | Summary |
|---|---|---|
| Name | 1 |
04 Members
要求两个张量形状一致
Wrap(
Double(), Int32())由计算结果的一维数组与形状包装出新的张量
MapUnary(Tensor,
Func(Of Double, Double))逐元素一元运算的通用标量实现
逐元素二元运算的通用标量实现
Heaviside(Tensor)
阶跃函数(Heaviside): 大于 0 的元素取 1, 否则取 0
ConvOutSize(
Int32, Int32, Int32, Int32)卷积/池化的输出边长: (input + 2*padding - kernel) / stride + 1
RequireRank4(Tensor,
String)卷积/池化的入参形状校验
沿指定轴执行规约操作（通用实现）
沿指定轴找到极值的索引（argmax / argmin）
Name
Exp(Tensor)
Log(Tensor)
Sqrt(Tensor)
Square(Tensor)
Abs(Tensor)
Sin(Tensor)
Cos(Tensor)
Tanh(Tensor)
Sigmoid(Tensor)
Negate(Tensor)
Reciprocal(Tensor)
Relu(Tensor)
LeakyRelu(Tensor,
Double)Elu(Tensor,
Double)Gelu(Tensor)
Swish(Tensor)
AddScalar(Tensor,
Double)MultiplyScalar(Tensor,
Double)DivideScalar(Tensor,
Double)Pow(Tensor,
Double)Transpose(Tensor)
Conv2DBackwardBias(Tensor)
SumAll(Tensor)
MeanAll(Tensor)
Prod(Tensor,
Nullable(Of Int32))StdDev(Tensor)
L2Norm(Tensor)
ArgMax(Tensor,
Nullable(Of Int32))ArgMin(Tensor,
Nullable(Of Int32))Softmax(Tensor,
Int32)LogSoftmax(Tensor,
Int32)L2Loss(Tensor)