List of all items

Structs

• attention::AttentionParams
• attention::FusedTransformer
• attention::LayerWeights
• attention::TransformerConfig
• attention::cpu::transformer::CpuKvCache
• backend::AttnConfig
• backend::KvBf16
• backend::KvCache
• backend::KvCacheQuant
• backend::KvFp16
• backend::KvFp8
• backend::KvInt8
• backend::MoeRouting
• backend::QuantWeights
• backend::cpu::CpuBackend
• backend::cpu::CpuGptqStore
• backend::reference::ReferenceDeviceBuffer
• backend::reference::ReferenceDeviceCommand
• backend::reference::ReferenceDeviceFence
• backend::reference::ReferenceDeviceRuntime
• backend::reference::ReferenceDeviceRuntimeError
• backend::reference::ReferenceDeviceRuntimeSnapshot
• backend::reference::ReferenceDeviceStream
• backend::reference::ReferenceVNextComposition
• backend::timer::CpuTimer
• linear::LinearMetadata
• marlin_repack::Fp8MarlinWeight
• native_ops::CompiledFa2NativeOperatorArtifact
• native_ops::CompiledNativeOperatorArtifact
• native_ops::NativeOperatorArtifactSpec
• native_ops::NativeOperatorRuntimeSelection
• quant_linear::cpu_dequant::CpuGptqLinear
• quant_linear::cpu_gguf::CpuGgufLinear
• quant_linear::cpu_marlin_stack::CpuMarlinExpertStack

Enums

• backend::GgufQuantType
• backend::QuantKind
• backend::ReduceOp
• backend::SrcDtype
• backend::buffer::CpuBuf
• backend::cpu::CpuQuantStore
• backend::dtype::Dtype
• linear::LinearProjectionRole
• marlin_repack::Fp8MarlinPrepareError

Traits

• backend::Backend
• backend::BackendCollective
• backend::BackendGraph
• backend::BackendInt8KvOps
• backend::BackendKvDtype
• backend::BackendMoeFused
• backend::BackendPagedKv
• backend::BackendQuantGguf
• backend::BackendQuantMarlin
• backend::KvDtypeKind
• backend::KvLayer
• backend::LlmBackend
• backend::MoeLlmBackend
• backend::QuantLlmBackend
• backend::dtype::HostDtype
• backend::timer::BackendTimer
• linear::Linear
• marlin_expert_stack::MarlinExpertStack
• stacked_expert::StackedExpertGgufLinear

Functions

• attention::attention
• attention::attention_cpu
• attention::cpu::fused_attention
• attention::cpu::gemm_at_b
• attention::cpu::softmax_inplace
• attention::cpu::transformer::cpu_layer_forward
• backend::reference::reference_vnext_capabilities
• backend::reference::reference_vnext_operation_registry
• backend::timer::finish_probe_timer
• backend::timer::finish_probe_timer_traced
• backend::timer::start_probe_timer_if
• configure_native_profile_sink
• marlin_fp8_materializer::marlin_fp8_weight_materializer
• marlin_repack::fp8_marlin_shape_supported
• marlin_repack::permute_gptq_qweight_rows
• marlin_repack::prepare_f16_weight_for_fp8_marlin
• marlin_repack::repack_gptq_to_marlin
• marlin_repack::repack_gptq_to_marlin_bytes_into
• marlin_repack::repack_scales_to_marlin
• moe_host::compute_ids_tpe
• native_ops::compiled_fa2_native_operator_artifact
• native_ops::compiled_fa2_native_operator_artifact_linked
• native_ops::compiled_fa2_native_operator_artifact_state
• native_ops::compiled_native_operator_artifacts
• native_ops::resolve_cuda_fa2_native_operator
• native_ops::resolve_native_operator_artifact
• native_ops::validate_compiled_native_operator_provider_catalog

Type Aliases

• attention::GpuBuffer

Constants

• backend::MAX_LAYERS_FOR_GRAPH
• backend::reference::REFERENCE_DENSE_SAFETENSORS_FORMAT_ID
• marlin_fp8_materializer::MARLIN_FP8_CAPABILITY_ID
• marlin_fp8_materializer::MARLIN_FP8_QUANTIZATION_FORMAT_ID
• marlin_fp8_materializer::MARLIN_FP8_WEIGHT_FORMAT_ID
• marlin_fp8_materializer::MARLIN_FP8_WEIGHT_LAYOUT_ID
• marlin_fp8_materializer::MARLIN_FP8_WEIGHT_MATERIALIZER_ID
• native_ops::CUDA_NATIVE_SOURCE_BUNDLE_ID
• native_ops::FA2_NATIVE_OPERATOR