pulsatrix::ReluModule Class Reference

y = max(x, 0), elementwise. No parameters, no parameter gradients. More...

#include <relu_module.hpp>

Inheritance diagram for pulsatrix::ReluModule:

Collaboration diagram for pulsatrix::ReluModule:

Public Member Functions
	ReluModule (DeviceBackend *backend, DeviceType device)
	Constructs a ReLU module.
	ReluModule (DeviceBackend *backend)
	As above, on backend's own device (backend->device()).
Tensor	backward (const Tensor &grad_output) override
	Computes the gradient w.r.t. this module's input.
OpType	op_type () const override
	Activation per charter's closed OpType set.
Tensor	propagate_relevance (const Tensor &relevance_out, const LRPRuleConfig &config) override
	Pass-through LRP relevance propagation.
bool	supports_lrp_rule (LRPRule) const override
	Pass-through relevance is the same under every rule: supports all of them.
std::optional< DeviceType >	compute_device () const override
	Where this layer computes, so forward() rejects an input on another device (FND-8).
Public Member Functions inherited from pulsatrix::Module
virtual	~Module ()=default
Tensor	forward (const Tensor &input)
	Runs this module's forward computation.
std::pair< Tensor, NodeId >	forward_traced (const Tensor &input, NodeId input_node, ComputationGraph &graph, Autograd &autograd)
	Runs forward() while also registering a ComputationGraph node (tagged with this module's op_type(), parented to input_node) and wiring an Autograd backward function that reuses this module's own backward() – the opt-in traced/explainable path, per Phase 2 Mission 0.
virtual std::vector< NamedParamRef >	named_parameters ()
	This module's trainable parameters, each with its hierarchical name – the one place a module declares its parameters (roadmap FND-1).
virtual std::vector< ParamRef >	parameters ()
	This module's trainable parameters and their gradients, for an optimizer to update uniformly across module types.
void	set_requires_grad (bool requires_grad, const std::string &prefix="")
	Freezes (false) or unfreezes (true) parameters by name (roadmap FND-2).
virtual void	set_training (bool training)
	Sets this module's training/eval mode. Defaults to training (matches every mainstream framework's Module default).
bool	is_training () const
	Whether this module is currently in training mode.

Protected Member Functions
Tensor	forward_impl (const Tensor &input) override
	The actual forward computation. Called by forward() after precondition checks.

Detailed Description

y = max(x, 0), elementwise. No parameters, no parameter gradients.

Constructor & Destructor Documentation

ReluModule() [1/2]

pulsatrix::ReluModule::ReluModule	(	DeviceBackend *	backend,
		DeviceType	device
	)

Constructs a ReLU module.

Parameters

backend	Backend to compute through. Not owned; must outlive this module.
device	Which device last_input_ is initially tagged as. forward_impl()'s output is tagged with the actual input tensor's device on every call (not this constructor argument), since ReLU has no parameters of its own to anchor a fixed "module device" the way LinearModule's weight_ does – see campaign_exai_dl_library_phase1_5_cuda_backend.md's Mission 3.

ReluModule() [2/2]

pulsatrix::ReluModule::ReluModule ( DeviceBackend *  backend )

explicit

As above, on backend's own device (backend->device()).

Member Function Documentation

backward()

Tensor pulsatrix::ReluModule::backward ( const Tensor &  grad_output )

overridevirtual

Computes the gradient w.r.t. this module's input.

Parameters

grad_output

Gradient w.r.t. this module's output. Must match the shape of the most recent forward() call's output.

Returns

Gradient w.r.t. this module's input: grad_output where input > 0, else 0.

Note

x == 0 is treated as blocked (project convention – ReLU's subgradient at 0 is technically any value in [0,1]; this codebase picks 0, matching forward's own x > 0 threshold for max(x, 0)).

Must be called after forward() – uses the input cached from that call.

Device-generic: runs on Cpu, Cuda or Hip tensors (GPU-native-kernels Mission 1).

Implements pulsatrix::Module.

compute_device()

std::optional< DeviceType > pulsatrix::ReluModule::compute_device ( ) const

inlineoverridevirtual

Where this layer computes, so forward() rejects an input on another device (FND-8).

Reimplemented from pulsatrix::Module.

forward_impl()

Tensor pulsatrix::ReluModule::forward_impl ( const Tensor &  input )

overrideprotectedvirtual

The actual forward computation. Called by forward() after precondition checks.

Implements pulsatrix::Module.

op_type()

OpType pulsatrix::ReluModule::op_type ( ) const

inlineoverridevirtual

Activation per charter's closed OpType set.

Implements pulsatrix::Module.

propagate_relevance()

Tensor pulsatrix::ReluModule::propagate_relevance ( const Tensor &  relevance_out, const LRPRuleConfig &  config  )

overridevirtual

Pass-through LRP relevance propagation.

Note

Not a placeholder – this IS the correct, standard rule for a pointwise nonlinearity. LRP rules are defined across weighted (Linear/Conv) connections; activation functions pass relevance through unchanged (Montavon et al. 2019; cross-checked against xai_context.aDNA's technique_lrp.md, whose own LRP-0/eps formulation operates on post-nonlinearity activations without giving the nonlinearity itself a separate redistribution rule).

Implements pulsatrix::Module.

supports_lrp_rule()

bool pulsatrix::ReluModule::supports_lrp_rule ( LRPRule  ) const

inlineoverridevirtual

Pass-through relevance is the same under every rule: supports all of them.

Reimplemented from pulsatrix::Module.

---

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

• include/pulsatrix/relu_module.hpp