Perform InnerProduct/FullyConnected/Dense forward calulations, allow fusing bias and activation into same GPU kernel. More...

Public Member Functions
virtual void	enqueue (Tensor &x, Tensor &w, Tensor *bias, Tensor &y, ExecutionContext const &e)=0

Static Public Member Functions
static std::unique_ptr< IPForward >	create (Context &ctx, IPSettings const &config, bool bias, StandardActivations activation=StandardActivations::identity)
	Create optimal object for innter product calculation. More...

Detailed Description

Perform InnerProduct/FullyConnected/Dense forward calulations, allow fusing bias and activation into same GPU kernel.

Member Function Documentation

static std::unique_ptr<IPForward> dlprim::core::IPForward::create	(	Context &	ctx,
		IPSettings const &	config,
		bool	bias,
		StandardActivations	activation = `StandardActivations::identity`
	)

static

Create optimal object for innter product calculation.

config - IP Settings, bias - apply bias activation - apply activation

The documentation for this class was generated from the following file: