Device-templated RoPE component. More...

Inheritance diagram for Mila::Dnn::Rope< TDeviceType, TPrecision >:

Public Types
using	ComponentBase = Component<TDeviceType, TPrecision>
using	MR = typename DeviceTypeTraits<TDeviceType>::memory_resource
using	TensorType = Tensor<TPrecision, MR>

Public Member Functions
	Rope (const std::string &name, const RopeConfig &config, std::optional< DeviceId > device_id=std::nullopt)
std::pair< TensorType &, TensorType & >	backward (TensorType &grad_Q, TensorType &grad_K)
	Backpropagate gradients through RoPE.
void	decode (TensorType &Q, TensorType &K, dim_t position)
	Single-token decode with explicit position.
void	forward (TensorType &Q, TensorType &K)
	Apply rotary position embeddings to Q and K in-place.
DeviceId	getDeviceId () const override
	Get the compute device id associated with this component.
std::vector< ITensor * >	getGradients () const override
	Return non-owning pointers to parameter gradient tensors.
MemoryStats	getMemoryStats () const override
	Return the current memory allocation breakdown for this component.
std::vector< ITensor * >	getParameters () const override
	Return non-owning pointers to parameter tensors.
MemoryStats	getRequiredMemory (const BuildContext &context) const override
	What onBuilding() would allocate for this context, without allocating.
const ComponentType	getType () const override
	Get the component type identifier.
dim_t	parameterCount () const override
	Return number of trainable parameters.
void	prefill (TensorType &Q, TensorType &K, dim_t position_offset)
	Apply rotary position embeddings with an explicit position offset.
void	save_ (ModelArchive &, SerializationMode) const override
void	synchronize () override
std::string	toString () const override
	Produce a short, human-readable description of the component.
void	zeroGradients () override
	Clear all model-owned gradients for this component.
Public Member Functions inherited from Mila::Dnn::Component< TDeviceType, TPrecision >
	Component (const std::string &name)
	Construct component with required name identifier.
virtual void	build (const BuildContext &context) final
	Build the component with the provided BuildContext (canonical overload).
const std::string	getName () const
	Get the component's name identifier.
virtual std::vector< std::string >	getParameterNames () const
	List all available parameter names for this component.
TrainingMode	getTrainingMode () const noexcept
	The current runtime behavioral mode of this Component.
virtual bool	isBuilt () const final
	Returns true if build() has completed successfully.
virtual void	load_ (ModelArchive &archive, SerializationMode mode)
	Restore this component's parameters from its archive scope.
virtual void	loadParameter (const std::string &, const Serialization::ITensorBlob &)
	Load a parameter from serialized tensor data.
virtual void	requireSerializableParameters () const
	Verify this component can serialize whatever parameters it owns.
virtual void	saveFlatTensors (Serialization::SafeTensorsWriter &writer, const std::string &prefix, Serialization::TensorSavePass pass) const
	Drive this component's tensors through one pass of a flat safetensors save.
void	setTrainingMode (TrainingMode mode)
	Set the runtime behavioral mode for this Component.

Protected Member Functions
void	onBuilding (const BuildContext &build_context) override
	Hook invoked by build() to allocate component buffers.
void	onExecutionContextSet () override
	Lifecycle hook: Called immediately after ExecutionContext is set.
void	onTrainingModeChanging (TrainingMode) override
	Hook called before TrainingMode transitions.
std::pair< shape_t, shape_t >	resolveRotatedShapes (const BuildContext &build_context) const
	The Q and K shapes this context implies.
Protected Member Functions inherited from Mila::Dnn::Component< TDeviceType, TPrecision >
IExecutionContext *	getExecutionContext () const
	Get the shared execution context.
bool	hasExecutionContext () const noexcept
	Check if execution context has been set.
template<TensorDataType TParameterPrecision, typename TMemoryResource>
void	loadParameterFromBlob (const std::string &param_name, const Serialization::ITensorBlob &blob, Tensor< TParameterPrecision, TMemoryResource > &target, const shape_t &expected_shape)
	Load a tensor blob into a parameter tensor with validation.
template<TensorDataType TParameterPrecision, typename TMemoryResource>
void	saveParameterToArchive (ModelArchive &archive, const std::string &parameter_name, const Tensor< TParameterPrecision, TMemoryResource > &parameter) const
	Write one parameter tensor into the archive under "tensors/<name>".
template<TensorDataType TParameterPrecision, typename TMemoryResource>
void	saveParameterToWriter (Serialization::SafeTensorsWriter &writer, const std::string &flat_name, const Tensor< TParameterPrecision, TMemoryResource > &parameter, Serialization::TensorSavePass pass) const
	Drive one parameter through one pass of a flat safetensors save.
void	setExecutionContext (IExecutionContext *context)
	Set the execution context for this component.

Additional Inherited Members
Static Public Member Functions inherited from Mila::Dnn::Component< TDeviceType, TPrecision >
static constexpr DeviceType	getDeviceType ()
	Compile-time device type for this component instance.
static constexpr TensorDataType	getPrecision () noexcept
	Compile-time tensor precision for this component instance.
Protected Types inherited from Mila::Dnn::Component< TDeviceType, TPrecision >
using	HostStagingMemoryResource
	Host memory a device-resident parameter stages through.
Protected Attributes inherited from Mila::Dnn::Component< TDeviceType, TPrecision >
BuildContext	build_context_ { shape_t{ 1 }, RuntimeMode::Training }
	The BuildContext stored at build time.

Detailed Description

template<DeviceType TDeviceType, TensorDataType TPrecision>
requires PrecisionSupportedOnDevice<TPrecision, TDeviceType>
class Mila::Dnn::Rope< TDeviceType, TPrecision >

Device-templated RoPE component.

Rotates Q and K in-place using a PairedOperation backend registered as "RopeOp". The component owns no forward output buffers – rotation writes directly back into the caller-provided tensors (typically views into a fused QKV projection buffer).

Supports three dispatch modes: forward(Q, K) – training, positions 0..T-1 prefill(Q, K, position_offset) – chunked prefill, positions offset..offset+T-1 decode(Q, K, position) – single-token KV-cache decode

Positional dispatch (prefill/decode) is available when the backend implements IPositionalPairedOp. The interface pointer is cached at operation creation via dynamic_cast.

This component has no trainable parameters.

Member Function Documentation

◆ backward()

template<DeviceType TDeviceType, TensorDataType TPrecision>

std::pair< TensorType &, TensorType & > Mila::Dnn::Rope< TDeviceType, TPrecision >::backward	(	TensorType &	grad_Q,
		TensorType &	grad_K )

inline

Backpropagate gradients through RoPE.

RoPE is an orthogonal rotation (R^T = R^{-1}), so input gradients are the upstream gradients rotated by the transpose (negative) angles.

Parameters

grad_Q	Upstream gradient w.r.t. rotated Q.
grad_K	Upstream gradient w.r.t. rotated K.

Returns: Pair of references: (grad_Q_in, grad_K_in).

◆ decode()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::decode	(	TensorType &	Q,
		TensorType &	K,
		dim_t	position )

inline

Single-token decode with explicit position.

Rotates Q and K in-place using the cos/sin cache row at position. Required for KV-cache autoregressive generation where T=1.

Parameters

Q	Query tensor [B, 1, n_heads * head_dim]. Mutated in-place.
K	Key tensor [B, 1, n_kv_heads * head_dim]. Mutated in-place.
position	Absolute position of the token in the full sequence.

◆ forward()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::forward	(	TensorType &	Q,
		TensorType &	K )

inline

Apply rotary position embeddings to Q and K in-place.

Uses implicit positions 0..T-1. Suitable for training forward passes.

Parameters

Q	Query tensor [B, T, n_heads * head_dim]. Mutated in-place.
K	Key tensor [B, T, n_kv_heads * head_dim]. Mutated in-place.

◆ getDeviceId()

template<DeviceType TDeviceType, TensorDataType TPrecision>

DeviceId Mila::Dnn::Rope< TDeviceType, TPrecision >::getDeviceId ( ) const

inlineoverridevirtual

Get the compute device id associated with this component.

Must return the device on which parameters and operations execute.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ getGradients()

template<DeviceType TDeviceType, TensorDataType TPrecision>

std::vector< ITensor * > Mila::Dnn::Rope< TDeviceType, TPrecision >::getGradients ( ) const

inlineoverridevirtual

Return non-owning pointers to parameter gradient tensors.

Gradient buffers are allocated only when the component is built in training mode, so a component built for inference returns an empty vector. Stateless components return empty in either mode. This is the accessor counterpart to getParameters() and does not throw on mode.

Returns: Vector of gradient pointers; empty when built for inference.

Exceptions

std::runtime_error if called before the component has been built.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ getMemoryStats()

template<DeviceType TDeviceType, TensorDataType TPrecision>

MemoryStats Mila::Dnn::Rope< TDeviceType, TPrecision >::getMemoryStats ( ) const

inlineoverridevirtual

Return the current memory allocation breakdown for this component.

Reflects allocations at the moment of the call. The returned stats naturally track the component lifecycle:

After construction – nothing; construction allocates none After build( Inference ) – parameters + T=1 state buffers After build( Training ) – parameters + T=full state buffers After setTrainingMode( Train ) – parameters + state + gradients

For CompositeComponent and Network, the returned stats are the recursive aggregate of all child components.

May be called at any time – no lifecycle preconditions.

Returns: MemoryStats reflecting current allocations.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ getParameters()

template<DeviceType TDeviceType, TensorDataType TPrecision>

std::vector< ITensor * > Mila::Dnn::Rope< TDeviceType, TPrecision >::getParameters ( ) const

inlineoverridevirtual

Return non-owning pointers to parameter tensors.

The returned tensor pointers remain valid for the lifetime of the component. Order should be canonical (weights before biases).

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ getRequiredMemory()

template<DeviceType TDeviceType, TensorDataType TPrecision>

MemoryStats Mila::Dnn::Rope< TDeviceType, TPrecision >::getRequiredMemory ( const BuildContext & context ) const

inlineoverridevirtual

What onBuilding() would allocate for this context, without allocating.

NOT additive across layers. The cos/sin caches are process-wide and shared through RopeCacheRegistry, so the operation reports what one owner pays and a transformer summing its layers must deduplicate by cache key. See CudaRopeOp::getRequiredStateMemorySize and MemoryFootprint.md section 6.2.

Reimplemented from Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ getType()

template<DeviceType TDeviceType, TensorDataType TPrecision>

const ComponentType Mila::Dnn::Rope< TDeviceType, TPrecision >::getType ( ) const

inlineoverridevirtual

Get the component type identifier.

Used for serialization and runtime type identification.

Returns: Component type enum value.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ onBuilding()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::onBuilding ( const BuildContext & build_context )

inlineoverrideprotectedvirtual

Hook invoked by build() to allocate component buffers.

Receives the stored BuildContext. Implementations must use config.allocationSeqLen() when sizing output buffers – this is the single call that makes Inference and Training allocate the correct buffer sizes automatically without per-component logic.

// Example -- Linear component:
shape_t out_shape =
{
    config.batchSize(),
    config.allocationSeqLen(),   // 1 for Inference, T for Training
    config_.getOutputFeatures()
};
output_ = std::make_unique<TensorType>( device, out_shape,
    this->getName() + ".output" );

The default implementation forwards to the legacy onBuilding( const shape_t& ) overload for backwards compatibility. New components should override this overload directly.

Note: Do not call build() or onBuilding() from within this hook.; Implementations should either succeed fully or leave no partial state, as a failed build() may be retried.

Takes the build-time configuration; use its allocationSeqLen() to obtain the correct output buffer sequence dimension.

Reimplemented from Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ onExecutionContextSet()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::onExecutionContextSet ( )

inlineoverrideprotectedvirtual

Lifecycle hook: Called immediately after ExecutionContext is set.

Override this to perform initialization that requires a valid ExecutionContext. At the time this is called, getExecutionContext() is guaranteed to return a valid context.

Common uses:

Composite components: Create and configure child components.
Device resource allocation: Query device capabilities.

Default implementation does nothing.

Exceptions

Any	exception thrown will cause setExecutionContext() to fail and restore the component to a "context not set" state.

Reimplemented from Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ onTrainingModeChanging()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::onTrainingModeChanging ( TrainingMode )

inlineoverrideprotectedvirtual

Hook called before TrainingMode transitions.

Called by setTrainingMode() after validation and lock acquisition, before the internal state is updated. Derived classes override to respond to the transition – e.g. zeroing gradient buffers on transition to Eval, or re-enabling dropout on transition to Training.

The default implementation is a no-op.

Takes the incoming TrainingMode.

Reimplemented from Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ parameterCount()

template<DeviceType TDeviceType, TensorDataType TPrecision>

dim_t Mila::Dnn::Rope< TDeviceType, TPrecision >::parameterCount ( ) const

inlineoverridevirtual

Return number of trainable parameters.

For leaf components this is the element count of owned parameter tensors. CompositeComponent and Network implementations should return the recursive aggregate across all children.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ prefill()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::prefill	(	TensorType &	Q,
		TensorType &	K,
		dim_t	position_offset )

inline

Apply rotary position embeddings with an explicit position offset.

Each token at chunk-local position t is rotated using the cos/sin cache row at absolute position (t + position_offset). Required for chunked prefill where successive chunks use increasing offsets.

Parameters

Q	Query tensor [B, T, n_heads * head_dim]. Mutated in-place.
K	Key tensor [B, T, n_kv_heads * head_dim]. Mutated in-place.
position_offset	Absolute position of the first token in this chunk.

◆ resolveRotatedShapes()

template<DeviceType TDeviceType, TensorDataType TPrecision>

std::pair< shape_t, shape_t > Mila::Dnn::Rope< TDeviceType, TPrecision >::resolveRotatedShapes ( const BuildContext & build_context ) const

inlineprotected

The Q and K shapes this context implies.

Shared with getRequiredMemory(), which runs before onBuilding() has assigned q_shape_/k_shape_ and would otherwise read them unset.

◆ save_()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::save_	(	ModelArchive &	,
		SerializationMode	) const

inlineoverridevirtual

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ synchronize()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::synchronize ( )

inlineoverridevirtual

         @brief Convenience accessor -- true if currently in Eval mode.

         Equivalent to getTrainingMode() == TrainingMode::Eval.
         Valid for both RuntimeMode::Inference and RuntimeMode::Training
         built components.

         @return true if in Eval mode.
        &zwj;/

bool isEvalMode() const noexcept { return getTrainingMode() == TrainingMode::Eval; }

    RuntimeMode getRuntimeMode() const noexcept
    {
        return build_context_.getRuntimeMode();
    }

    bool isInferenceMode() const noexcept
    {
        return build_context_.isInferenceMode();
    }

    bool isTrainingMode() const noexcept
    {
        return build_context_.isTrainingMode();
    }

====================================================================

Synchronization

    /**
       @brief Wait for outstanding device work submitted by this component.

       On CPU this may be a no-op. Use to ensure results are visible to
       the host or to measure synchronous timings.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ toString()

template<DeviceType TDeviceType, TensorDataType TPrecision>

std::string Mila::Dnn::Rope< TDeviceType, TPrecision >::toString ( ) const

inlineoverridevirtual

Produce a short, human-readable description of the component.

Implementations should keep output concise and avoid throwing.

Implements Mila::Dnn::Component< TDeviceType, TPrecision >.

◆ zeroGradients()

template<DeviceType TDeviceType, TensorDataType TPrecision>

void Mila::Dnn::Rope< TDeviceType, TPrecision >::zeroGradients ( )

inlineoverridevirtual

Clear all model-owned gradients for this component.

Default implementation is a no-op. Composite components should override to recurse to children. Leaf components should override to zero their parameter and activation gradients using device-aware helpers.

Reimplemented from Mila::Dnn::Component< TDeviceType, TPrecision >.

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

Mila/Src/Dnn/Components/Encodings/Rope/Rope.ixx

Public Types

Public Member Functions

Protected Member Functions

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ backward()

◆ decode()

◆ forward()

◆ getDeviceId()

◆ getGradients()

◆ getMemoryStats()

◆ getParameters()

◆ getRequiredMemory()

◆ getType()

◆ onBuilding()

◆ onExecutionContextSet()

◆ onTrainingModeChanging()

◆ parameterCount()

◆ prefill()

◆ resolveRotatedShapes()

◆ save_()

◆ synchronize()

Synchronization

◆ toString()

◆ zeroGradients()