Splits num_channels into num_groups equal-size groups; each group's mean/std is computed over every (channel-in-group, H, W) element jointly, per batch row n, then y_{n,c,h,w} = gamma_c * (x_{n,c,h,w} - mu_{n,g})/std_{n,g} + beta_c, gamma/beta per-channel (shape (num_channels,), not per-group, not per-batch-row). Batched – input/output are rank-4 (N, channels, H, W), migrated from the original unbatched (rank-3) scope by campaign_exai_dl_library_batch_dimension_support, matching Conv2DModule's own (not-yet-migrated) rank-3 convention plus a leading batch dim. H/W are not fixed at construction (only num_groups/num_channels are), so this module accepts any spatial size at forward() time, exactly like Conv2DModule does. More...

#include <group_norm_module.hpp>

Inheritance diagram for pulsatrix::GroupNormModule:

[legend]

Collaboration diagram for pulsatrix::GroupNormModule:

[legend]

Public Member Functions
	GroupNormModule (int64_t num_groups, int64_t num_channels, DeviceBackend *backend, DeviceType device, float eps=1e-6f)
	Constructs a GroupNorm layer with zero-initialized gamma and beta.

	GroupNormModule (int64_t num_groups, int64_t num_channels, DeviceBackend *backend)
	On backend's own device (backend->device()), default eps. Previously the device defaulted to Cpu regardless of backend (GPU-native-kernels Mission 0).

Tensor	backward (const Tensor &grad_output) override
	Computes the gradient w.r.t. this module's input, and accumulates gamma's/ beta's gradients internally.

OpType	op_type () const override
	Normalization per charter's closed OpType set.

void	set_gamma (std::initializer_list< float > values)
	Overwrites the per-channel gamma buffer – test/initialization use only.

void	set_beta (std::initializer_list< float > values)
	Overwrites the per-channel beta buffer – test/initialization use only.

void	set_gamma (const std::vector< float > &values)
	Vector overload for runtime-sized sources – see Tensor's own vector ctor.

void	set_beta (const std::vector< float > &values)
	Vector overload for runtime-sized sources – see Tensor's own vector ctor.

const Tensor &	gamma () const

const Tensor &	beta () const

const Tensor &	gamma_grad () const

const Tensor &	beta_grad () const

Tensor	propagate_relevance (const Tensor &relevance_out, const LRPRuleConfig &config) override
	Identity-rule LRP relevance propagation (AttnLRP, Achtibat et al. 2024).

std::vector< NamedParamRef >	named_parameters () override
	This module's trainable parameters, each with its hierarchical name – the one place a module declares its parameters (roadmap FND-1).

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 bool	supports_lrp_rule (LRPRule rule) const
	Whether propagate_relevance() implements `rule` (no silent fallback: callers such as ExplainerContext::relevance_pass() throw rather than run a module on a rule it does not implement).

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

Splits num_channels into num_groups equal-size groups; each group's mean/std is computed over every (channel-in-group, H, W) element jointly, per batch row n, then y_{n,c,h,w} = gamma_c * (x_{n,c,h,w} - mu_{n,g})/std_{n,g} + beta_c, gamma/beta per-channel (shape (num_channels,), not per-group, not per-batch-row). Batched – input/output are rank-4 (N, channels, H, W), migrated from the original unbatched (rank-3) scope by campaign_exai_dl_library_batch_dimension_support, matching Conv2DModule's own (not-yet-migrated) rank-3 convention plus a leading batch dim. H/W are not fixed at construction (only num_groups/num_channels are), so this module accepts any spatial size at forward() time, exactly like Conv2DModule does.

Note: propagate_relevance is an identity pass-through, cited to AttnLRP's normalization-layer treatment (Achtibat et al. 2024) – same rule, same citation, as RMSNormModule/LayerNormModule; GroupNorm is architecturally the same "normalization" category the citation covers, just with a different statistic grouping. backward() is the real, undetached training gradient.

Constructor & Destructor Documentation

◆ GroupNormModule() [1/2]

pulsatrix::GroupNormModule::GroupNormModule	(	int64_t	num_groups,
		int64_t	num_channels,
		DeviceBackend *	backend,
		DeviceType	device,
		float	eps = `1e-6f`
	)

Constructs a GroupNorm layer with zero-initialized gamma and beta.

Parameters

num_groups	Number of groups to split num_channels into.
num_channels	Total channel count. Must be evenly divisible by num_groups.
backend	Backend to allocate/compute through. Not owned; must outlive this module.
device	Which device every internal Tensor member is tagged as. Defaults to Cpu.
eps	Stabilizer added inside the sqrt. Defaults to 1e-6, matching RMSNormModule/LayerNormModule's default for consistency within this codebase's normalization family.

Exceptions

std::invalid_argument if num_groups <= 0, num_channels <= 0, or num_channels is not evenly divisible by num_groups – external boundary (construction arguments can originate from Phase 5's Python bindings with no upstream validation), per cpp_tdd/context_tdd_adversarial_boundary_testing.md.

◆ GroupNormModule() [2/2]

pulsatrix::GroupNormModule::GroupNormModule	(	int64_t	num_groups,
		int64_t	num_channels,
		DeviceBackend *	backend
	)

On backend's own device (backend->device()), default eps. Previously the device defaulted to Cpu regardless of backend (GPU-native-kernels Mission 0).

Member Function Documentation

◆ backward()

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

overridevirtual

Computes the gradient w.r.t. this module's input, and accumulates gamma's/ beta's gradients internally.

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.

Exceptions

std::logic_error if forward() has never been called.

Note: Device-generic: runs on Cpu, Cuda or Hip tensors (GPU-native-kernels Mission 4).

Implements pulsatrix::Module.

◆ beta()

const Tensor & pulsatrix::GroupNormModule::beta ( ) const

inline

◆ beta_grad()

const Tensor & pulsatrix::GroupNormModule::beta_grad ( ) const

inline

◆ compute_device()

std::optional< DeviceType > pulsatrix::GroupNormModule::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::GroupNormModule::forward_impl ( const Tensor & input )

overrideprotectedvirtual

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

Implements pulsatrix::Module.

◆ gamma()

const Tensor & pulsatrix::GroupNormModule::gamma ( ) const

inline

◆ gamma_grad()

const Tensor & pulsatrix::GroupNormModule::gamma_grad ( ) const

inline

◆ named_parameters()

std::vector< NamedParamRef > pulsatrix::GroupNormModule::named_parameters ( )

inlineoverridevirtual

This module's trainable parameters, each with its hierarchical name – the one place a module declares its parameters (roadmap FND-1).

Returns: {name, {value, grad}} entries pointing directly at this module's own members, in a fixed order. Names are unique within the module tree. Default: empty (a parameterless module like ReluModule needs no override).

Note: Override this, not parameters(): saving, loading, freezing by name and optimizer parameter groups all key on these names.

Reimplemented from pulsatrix::Module.

◆ op_type()

OpType pulsatrix::GroupNormModule::op_type ( ) const

inlineoverridevirtual

Normalization per charter's closed OpType set.

Implements pulsatrix::Module.

◆ propagate_relevance()

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

overridevirtual

Identity-rule LRP relevance propagation (AttnLRP, Achtibat et al. 2024).

Parameters

relevance_out	Relevance at this module's output. Must match the shape of the most recent forward() call's output.
config	Unused – the identity rule has no tunable parameter.

Returns: relevance_out, unchanged – conservation holds trivially by construction.

Exceptions

std::logic_error	if forward() has never been called.
std::invalid_argument	if relevance_out's element count doesn't match the cached forward output's element count.

Implements pulsatrix::Module.

◆ set_beta() [1/2]

void pulsatrix::GroupNormModule::set_beta ( const std::vector< float > & values )

Vector overload for runtime-sized sources – see Tensor's own vector ctor.

◆ set_beta() [2/2]

void pulsatrix::GroupNormModule::set_beta ( std::initializer_list< float > values )

Overwrites the per-channel beta buffer – test/initialization use only.

◆ set_gamma() [1/2]

void pulsatrix::GroupNormModule::set_gamma ( const std::vector< float > & values )

Vector overload for runtime-sized sources – see Tensor's own vector ctor.

◆ set_gamma() [2/2]

void pulsatrix::GroupNormModule::set_gamma ( std::initializer_list< float > values )

Overwrites the per-channel gamma buffer – test/initialization use only.

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

include/pulsatrix/group_norm_module.hpp

Public Member Functions

Protected Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ GroupNormModule() [1/2]

◆ GroupNormModule() [2/2]

Member Function Documentation

◆ backward()

◆ beta()

◆ beta_grad()

◆ compute_device()

◆ forward_impl()

◆ gamma()

◆ gamma_grad()

◆ named_parameters()

◆ op_type()

◆ propagate_relevance()

◆ set_beta() [1/2]

◆ set_beta() [2/2]

◆ set_gamma() [1/2]

◆ set_gamma() [2/2]