torch.logsumexp
function logsumexp<S extends Shape, D extends DType, Dev extends DeviceType>(input: Tensor<S, D, Dev>, options: LogsumexpOptions & { dim: number }): Tensor<DynamicShape, D, Dev>function logsumexp<S extends Shape, D extends DType, Dev extends DeviceType>(input: Tensor<S, D, Dev>, dim: number, options?: LogsumexpOptions): Tensor<DynamicShape, D, Dev>Returns the log of summed exponentials along a dimension in a numerically stable way. Computes: log(sum(exp(x))) along the specified dimension.
Parameters
inputTensor<S, D, Dev>- The input tensor
optionsLogsumexpOptions & { dim: number }
Returns
Tensor<DynamicShape, D, Dev>– Tensor with logsumexp valuesExamples
const x = torch.tensor([1.0, 2.0, 3.0]);
torch.logsumexp(x, 0); // log(e^1 + e^2 + e^3)