Dnn.Components.QwenAttentionBlock Module Reference

Exported Modules
module	Dnn.Components.Gqa
module	Serialization.Mode
module	Dnn.TensorDataType
module	Compute.Device
module	Compute.DeviceId
module	Dnn.Components.QwenConfig
module	Compute.CpuMemoryResource
module	Dnn.Component
module	Dnn.ComponentType
module	Compute.IExecutionContext
module	Dnn.Components.AttentionOutputGate
module	Dnn.Components.Swiglu
module	Dnn.ITensor
module	Compute.ExecutionContext
module	Compute.DeviceType
module	Serialization.ModelArchive
module	Dnn.Quantization.Weight.Policies
module	Dnn.Components.RmsNorm
module	Dnn.Quantization.Weight.PrecisionPlan
module	Dnn.ActivationType
module	Dnn.CompositeComponent
module	Dnn.Quantization.KvCache.Policy
module	Compute.GqaState
module	Dnn.TensorDataTypeTraits
module	Dnn.TensorTypes
module	Compute.DeviceAllocation
module	Dnn.Components.Linear
module	Dnn.Tensor
module	Compute.Observation
module	Compute.DeviceTypeTraits
module	Compute.ExecutionContextFactory
module	Dnn.Components.Residual
module	Dnn.Components.Rope
module	Dnn.Components.ITransformerBlock
module	Dnn.TensorOps

Classes
struct	Mila::Dnn::QwenAttentionBlock< TDeviceType, TPrecision, TWeightPlan, TKvPolicy >::BlockBuildContexts
	The per-child build contexts and split-scratch geometry this block implies. More...
class	Mila::Dnn::QwenAttentionBlock< TDeviceType, TPrecision, TWeightPlan, TKvPolicy >
	One Qwen 3.8 full-attention decoder block. More...
struct	Mila::Dnn::QwenAttentionBlockWorkspace< TDeviceType, TPrecision >
	Transformer-owned shared activation workspace for QwenAttentionBlock (pooling). More...

Functions
template<DeviceType TDeviceType, TensorDataType TPrecision> requires PrecisionSupportedOnDevice<TPrecision, TDeviceType>
QwenAttentionBlockWorkspace< TDeviceType, TPrecision >	Mila::Dnn::makeQwenAttentionBlockWorkspace (const QwenConfig &config, DeviceId device, dim_t B, dim_t prefill_chunk, const std::string &name_prefix)
	Allocate the shared attention-block workspace.

Files
file	Mila/Src/Dnn/Components/Transformers/Qwen/Qwen.AttentionBlock.ixx
	Qwen 3.8 full-attention decoder block (inference: prefill + decode).