Struct cudnn::Cudnn [−] [src]

pub struct Cudnn {
    // some fields omitted
}

Provides a the high-level interface to CUDA's cuDNN.

Methods

impl Cudnn
[src]

fn new() -> Result<Cudnn, Error>

Initializes a new CUDA cuDNN context.

Make sure your current CUDA device is cuDNN enabled.

fn from_c(id: cudnnHandle_t) -> Cudnn

Initializes a new CUDA cuDNN Context from its C type.

fn id(&self) -> isize

Returns the id as isize.

fn id_c(&self) -> cudnnHandle_t

Returns the CUDA cuDNN Context as its C type.

fn version() -> usize

Returns the version of the CUDA cuDNN library.

fn sigmoid_forward(&self, src_desc: &TensorDescriptor, src_data: *const c_void, dest_desc: &TensorDescriptor, dest_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the forward Sigmoid Activation function.

Writes the result of the computation to dest_data.

fn sigmoid_backward(&self, src_desc: &TensorDescriptor, src_data: *const c_void, src_diff_desc: &TensorDescriptor, src_diff_data: *const c_void, dest_desc: TensorDescriptor, dest_data: *const c_void, dest_diff_desc: TensorDescriptor, dest_diff_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the backward Sigmoid Activation function.

Writes the result of the computation to dest_diff_data.

fn relu_forward(&self, src_desc: TensorDescriptor, src_data: *const c_void, dest_desc: TensorDescriptor, dest_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the forward Rectified Linear Activation function.

Writes the result of the computation to dest_data.

fn relu_backward(&self, src_desc: &TensorDescriptor, src_data: *const c_void, src_diff_desc: &TensorDescriptor, src_diff_data: *const c_void, dest_desc: TensorDescriptor, dest_data: *const c_void, dest_diff_desc: TensorDescriptor, dest_diff_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the backward Rectified Linear Activation function.

Writes the result of the computation to dest_diff_data.

fn tanh_forward(&self, src_desc: TensorDescriptor, src_data: *const c_void, dest_desc: TensorDescriptor, dest_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the forward Hyperbolic Tangent Activation function.

Writes the result of the computation to dest_data.

fn tanh_backward(&self, src_desc: &TensorDescriptor, src_data: *const c_void, src_diff_desc: &TensorDescriptor, src_diff_data: *const c_void, dest_desc: TensorDescriptor, dest_data: *const c_void, dest_diff_desc: TensorDescriptor, dest_diff_data: *mut c_void, scale: ScalParams) -> Result<(), Error>

Computes the backward Hyperbolic Tangent Activation function.

Writes the result of the computation to dest_diff_data.

Trait Implementations

impl Clone for Cudnn
[src]

fn clone(&self) -> Cudnn

Returns a copy of the value. Read more

fn clone_from(&mut self, source: &Self)
1.0.0

Performs copy-assignment from source. Read more

impl Debug for Cudnn
[src]

fn fmt(&self, __arg_0: &mut Formatter) -> Result

Formats the value using the given formatter.

impl Drop for Cudnn
[src]

fn drop(&mut self)

A method called when the value goes out of scope. Read more