Class CudaTensor (mod)

Functions for creating and operating on tensors stored on a nVidia CUDA GPU.

Inheritance

CudaTensor (mod)

Namespace: Tensor

Assembly: Tensor.dll

Syntax

module CudaTensor

Remarks

This module contains functions for creating tensors stored on a nVidia CUDA GPU. It further contains functions that only work with tensors stored on a nVidia CUDA GPU. Calling these functions with tensors stored on other devices will result in an System.InvalidOperationException.

The CUDA backend provides options that can be configured via Cfg.

Examples

 let x = CudaTensor.zeros [3L; 3L]  // x.Dev = CudaTensor.Dev

Fields

Name	Description
val Dev	Tensor device using CUDA GPU memory as data storage.

Methods

Name	Description
val arange: 'T -> 'T -> 'T -> Tensor<'T>	Creates a new vector filled with equaly spaced values using a specifed increment.
val counting: int64 -> Tensor<int64>	Creates a new vector filled with the integers from zero to the specified maximum.
val empty: int -> Tensor<'T>	Creates a new, empty tensor with the given number of dimensions.
val falses: int64 list -> Tensor<bool>	Creates a new boolean tensor filled with falses.
val filled: int64 list -> 'T -> Tensor<'T>	Creates a new tensor filled with the specified value.
val identity: int64 -> Tensor<'T>	Creates a new identity matrix.
val linspace: 'T -> 'T -> int64 -> Tensor<'T>	Creates a new vector of given size filled with equaly spaced values.
val ones: int64 list -> Tensor<'T>	Creates a new tensor filled with ones (1).
val scalar: 'T -> Tensor<'T>	Creates a new zero-dimensional (scalar) tensor with the specified value.
val transfer: Tensor<'T> -> Tensor<'T>	Transfers a tensor to the CUDA device.
val trues: int64 list -> Tensor<bool>	Creates a new boolean tensor filled with trues.
val usingPtr: CUdeviceptr -> SizeT -> TensorLayout -> Tensor<'T>	Creates a tensor from the given CUDA pointer, allocation size in bytes and layout.
val usingPtrAndType: CUdeviceptr -> SizeT -> Type -> TensorLayout -> ITensor	Creates a tensor from the given CUDA pointer, allocation size in bytes, type and layout.
val zeros: int64 list -> Tensor<'T>	Creates a new tensor filled with zeros (0).