List of all items

Structs

• circuit_exec::Boot
• circuit_exec::CircuitExec
• circuit_exec::TargetModules
• circuit_exec::bindings::AttnFacts
• circuit_exec::bindings::CircuitLayer
• circuit_exec::bindings::DraftBinding
• circuit_exec::bindings::DraftRows
• circuit_exec::bindings::GdnFacts
• circuit_exec::bindings::HeadBinding
• circuit_exec::bindings::RopeFacts
• circuit_exec::compile::Inputs
• circuit_exec::fixed::DraftFixed
• circuit_exec::fixed::Fixed
• circuit_exec::kernels::KernelTable
• circuit_exec::program::DraftPrograms
• circuit_exec::program::GdnState
• circuit_exec::program::Launch
• circuit_exec::program::Program
• circuit_exec::program::StepEnv
• circuit_exec::routes::RoutedProgram
• circuit_exec::verify_batch::Parts
• circuit_exec::verify_batch::VerifyBatch
• circuit_exec::verify_batch::VerifyBatchProgram
• drafter_context::DrafterContext
• forward::qwen3_5::FullAttentionLayer
• forward::qwen3_5::FullAttentionScratch
• forward::qwen3_5::LayerKvCache
• forward::qwen3_5::LinearAttentionLayer
• forward::qwen3_5::LinearAttentionScratch
• forward::qwen3_5::LinearAttentionState
• forward::qwen3_5::Qwen35ForwardConfig
• forward::qwen3_5::Qwen35Kernels
• layer::AttnLayerState
• layer::AttnMetadataDev
• layer::BatchedAttnMetadata
• layer::EmptyLayerState
• layer::ForwardContext
• layer::GdnCarryBinding
• layer::GdnPrefillBuffers
• layer::MidchunkCapture
• layer::SsmLayerState
• layers::W8a8Ctx
• layers::W8a8Ffn
• layers::W8a8Mixer
• layers::dense_ffn::DenseFfnLayer
• layers::dense_ffn::DenseFfnWeights
• layers::dense_ffn::DenseFfnWeightsBf16
• layers::dense_ffn::DenseFfnWeightsFp8
• layers::dense_ffn::DenseFfnWeightsQ2
• layers::dense_ffn::m16_tc::M16TcLevers
• layers::dense_ffn::m16_tc::oracle::M16TcDiff
• layers::dense_ffn::m16_tc::oracle::M16TcOutlier
• layers::ep_dispatch::EpRoutingTable
• layers::fp8_calibration::Fp8KvCalibration
• layers::fp8_calibration::Fp8KvWriteTarget
• layers::glm_vit::GlmVit
• layers::glm_vit::GlmVitBlock
• layers::glm_vit::GlmVitMerger
• layers::glm_vit::GlmVitScratch
• layers::glm_vit::init::GlmVitGeometry
• layers::moe::MoeLayer
• layers::mtp_head::MtpHead
• layers::mtp_head::MtpProposerState
• layers::mtp_multi::MultiModuleMtpHead
• layers::mtp_multi::MultiModuleMtpState
• layers::ngram_embed::NgramDims
• layers::ngram_embed::NgramEmbedding
• layers::ops::AttnFa128Kernels
• layers::ops::BaGatesPick
• layers::ops::CublasScope
• layers::ops::DecodeW8a8Plan
• layers::ops::DecodeW8a8Scratch
• layers::ops::DenseMmKernels
• layers::ops::DerivedWeights
• layers::ops::E4m3Saturation
• layers::ops::E4m3SaturationScope
• layers::ops::Fp8ActQuant
• layers::ops::Fp8GroupedGeometry
• layers::ops::Fp8GroupedSortOut
• layers::ops::Fp8GroupedW8a8Layout
• layers::ops::Fp8GroupedW8a8Rows
• layers::ops::Fp8QuantPick
• layers::ops::GdnCarrySizes
• layers::ops::GemmDispatch
• layers::ops::ModelLevers
• layers::ops::MoeCutlassDownHostTables
• layers::ops::MoeCutlassHostTables
• layers::ops::MoeE4m3Tile
• layers::ops::Nvfp4ExpertTables
• layers::ops::W8a8Kernels
• layers::ops::W8a8Scratch
• layers::ops::W8a8Weight
• layers::ops::lora_delta::LoraAttnWeights
• layers::ops::lora_delta::LoraFfnWeights
• layers::ops::lora_delta::LoraKernels
• layers::ops::lora_delta::LoraPair
• layers::ops::lora_delta::LoraRoute
• layers::ops::model_stats::DumpLatches
• layers::ops::model_stats::ModelStats
• layers::ops::model_stats::MoeUnionStats
• layers::ops::moe_lora_grouped::ExpertTables
• layers::ops::moe_lora_grouped::MoeExpertRoute
• layers::ops::target_defaults::Resolved
• layers::ops::target_defaults::TargetLevers
• layers::ops::token_overlay::OverlayKernels
• layers::ops::w4a4_proj::Nvfp4ActBuf
• layers::ops::w4a4_proj::W4a4Proj
• layers::ple::PleLayer
• layers::ple::PleSeqState
• layers::ple::PleWeights
• layers::ple::ids::PleIdDims
• layers::qsa::QsaIndexer
• layers::qsa::QsaSelection
• layers::qsa::QsaSeqState
• layers::qwen3_attention::CompressorWeights
• layers::qwen3_attention::Fp8TwinSet
• layers::qwen3_attention::HcHeadWeights
• layers::qwen3_attention::HcLowRank
• layers::qwen3_attention::HcSiteWeights
• layers::qwen3_attention::HcWeights
• layers::qwen3_attention::MlaWeights
• layers::qwen3_attention::innerq_driver::InnerQDriver
• layers::qwen3_attention::types::HcWeights
• layers::qwen3_attention::types::MlaWeights
• layers::qwen3_attention::types::Qwen3AttentionLayer
• layers::qwen3_ssm::Qwen3SsmLayer
• layers::qwen3_ssm::gdn_flags::GdnFlags
• layers::vision_encoder::MergerLayer
• layers::vision_encoder::ViTBlock
• layers::vision_encoder::VisionEncoder
• layers::vision_encoder::VisionScratch
• layers::w4a16_gemv_tiers::W4a16BatchmTiers
• lora::AdapterSlot
• lora::EmbedOverlay
• lora::ExpertLoraLayer
• lora::ExpertWork
• lora::LmHeadOverlay
• lora::LoraAdapterInput
• lora::LoraLayerWeights
• lora::LoraWeights
• lora::OverlayRaw
• lora::OverlayRawSlot
• lora::OverlayTensor
• lora::OverlayTensors
• lora::SlotView
• lora::TokenOverlaySet
• mtp_carry::AppendPlan
• mtp_carry::CarriedDrafter
• mtp_carry::StoreRange
• quant_format::CompressedTensorsFormat
• quant_format::Fp8BlockScaledFormat
• quant_format::IgnoreList
• quant_format::ModeloptFormat
• speculative::tree_shape::HedgeNode
• speculative::tree_shape::TreeDraft
• speculative::tree_shape::TreeRow
• speculative::tree_shape::TreeShape
• ssm_reserve::CopyTier
• ssm_reserve::DecodeRingDecision
• ssm_reserve::MarconiSlotDecision
• ssm_reserve::PoolCounts
• ssm_reserve::PoolPlan
• ssm_reserve::PoolShape
• ssm_reserve::VerifyCounts
• video_decode_ffmpeg::FfmpegPolicy
• video_preprocess::PreprocessedVideo
• vision_item::VisionItem
• weight_map::AttentionWeights
• weight_map::DenseExpertWeight
• weight_map::ExpertWeight
• weight_map::Fp8ExpertWeight
• weight_map::MoeWeights
• weight_map::MtpFp8Experts
• weight_map::MtpWeights
• weight_map::QuantExpertWeight
• weight_map::SsmWeights
• weight_map::model_a::ModelWeights
• weight_map::nemotron::NemotronExpertWeight
• weight_map::nemotron::NemotronMoeWeights
• weight_map::nemotron::NemotronSsmWeights
• weight_map::nvfp4_detect::QuantizeCtx
• weight_map::quantized::DenseWeight
• weight_map::quantized::Fp8DenseWeight
• weight_map::quantized::Fp8Weight
• weight_map::quantized::Fp8WeightTransposed
• weight_map::quantized::PackedQ2Weight
• weight_map::quantized::QuantizedWeight
• weight_map::ssm_qwen35::Nvfp4MoeCopies
• weight_map::ssm_qwen35::SsmWeightsQwen35

Enums

• circuit_exec::Fusions
• circuit_exec::bindings::BoundWeight
• circuit_exec::bindings::MixerFacts
• circuit_exec::bindings::WeightSlot
• circuit_exec::program::LaunchKind
• forward::qwen3_5::MetalKvDtype
• layer::MoeLoraRoute
• layers::ExpertQuantization
• layers::FfnComponent
• layers::MoeExpertTables
• layers::RowTiers
• layers::dense_ffn::FfnActivation
• layers::moe::GroupedRouting
• layers::mtp_head::MtpQuantization
• layers::ngram_embed::NgramTable
• layers::ops::BaGatesLogSlot
• layers::ops::Derivation
• layers::ops::F16ChainStates
• layers::ops::Fp8QuantLogSlot
• layers::ops::GdnScalarSpine
• layers::ops::HcVariant
• layers::ops::ResidentMat
• layers::ops::W8a8Scale
• layers::ops::dense_gemv_tc::DtcKind
• layers::ops::gemv_tc::TcKind
• layers::ops::target_defaults::Source
• layers::qwen3_attention::types::HeadGateActivation
• layers::vision_tower::VisionTower
• lora::AdapterAb
• lora::ExpertProj
• lora::LoraModule
• lora::LoraSites
• lora::LoraTarget
• lora::OverlayModule
• lora::OverlayTensorKind
• lora::VictimError
• mtp_carry::CarryOutcome
• speculative::ladder::MtpMaxSeqsSource
• ssm_reserve::PoolState
• ssm_reserve::SsmRollbackMode
• ssm_reserve::UnitSource
• video_decode_ffmpeg::Availability
• weight_map::LayerWeights
• weight_map::Nvfp4Variant
• weight_map::QuantWeight
• weight_map::nemotron::NemotronSsmQuant
• weight_map::quantized::WeightQuantFormat

Traits

• circuit_exec::bindings::CircuitBindings
• circuit_exec::program::DraftRunner
• forward::quant_weights::QuantWeights
• layer::LayerAuxState
• layer::LayerCapabilities
• layer::LayerGraphHooks
• layer::LayerSplitPrefill
• layer::LayerState
• layer::LayerWeightSetup
• layer::LayerWriteOnAccept
• layer::TransformerLayer
• quant_format::QuantFormat
• speculative::DraftProposer
• speculative::ProposerState

Functions

• circuit_exec::bindings::check_bindings
• circuit_exec::compile::compile
• circuit_exec::compile::layout
• circuit_exec::kernels::available_in
• circuit_exec::kernels::ptx_defines
• circuit_exec::policy::kv_dtype_name
• circuit_exec::policy::live_policy
• circuit_exec::policy::unmodelled_switches
• circuit_exec::routes::check_known
• circuit_exec::routes::holds
• circuit_exec::sources::checkpoint_shape
• circuit_exec::sources::instance
• circuit_exec::sources::instance_for
• circuit_exec::sources::lookup
• circuit_exec::sources::served_shape
• circuit_exec::sources::shape_drift
• circuit_exec::sources::sources
• circuit_exec::spec_key::spec_cost_plan_digests
• circuit_exec::verify_batch::arena_bytes
• circuit_exec::verify_batch::sizing_tables
• drafter_context::resolve
• drafter_context::resolve_from_env
• forward::qwen3_5::forward_full_attention
• forward::qwen3_5::forward_linear_attention
• layers::act_route
• layers::activation_quantization
• layers::any_fixed
• layers::dense_ffn::batch16_decode::ffn_batch16_enabled
• layers::dense_ffn::gateup_fused::ffn_gateup_fused
• layers::dense_ffn::m16_tc::m16_tc_levers
• layers::dense_ffn::m16_tc::oracle::bf16_ord
• layers::dense_ffn::m16_tc::oracle::compare_m16_tc_block
• layers::dense_ffn::m16_tc::oracle::m16_tc_acc_floor
• layers::dense_ffn::m16_tc::oracle::within_m16_tc_budget
• layers::dense_ffn::w8a8_prefill::ffn_w8a16_only
• layers::ep_dispatch::build_ep_routing_table
• layers::expert_quantization
• layers::family_fixed
• layers::fixed_act
• layers::fp8_calibration::dtype_runs_online_fp8_kv_calibration
• layers::fp8_calibration::graphs_ready_after_fp8_kv_cal
• layers::glm_vit::block_major_position_ids
• layers::glm_vit::build_axial_rope_tables
• layers::k64_kernel
• layers::k64_n64_kernel
• layers::k64_n64_wins
• layers::kernel_caps
• layers::moe::fp8_grouped_decode_shape_ok
• layers::moe::moe_expert_fp8_act
• layers::moe::nvfp4_grouped_decode_shape_ok
• layers::moe::set_moe_expert_fp8_act
• layers::moe_expert_tables
• layers::moe_grouped_decode_decide
• layers::moe_grouped_decode_enabled
• layers::moe_grouped_decode_for
• layers::moe_grouped_decode_forced
• layers::moe_grouped_decode_min_rows
• layers::mtp_head::mtp_drafter_prefill_enabled
• layers::mtp_head::mtp_experts_nvfp4
• layers::mtp_head::set_mtp_experts_nvfp4_from_cli
• layers::mtp_meta::mtp_attn_meta_batch_dev
• layers::mtp_meta::mtp_attn_meta_dev
• layers::mtp_meta::pack_mtp_attn_meta
• layers::ngram_embed::ngram_ids
• layers::ops::allow_e4m3_saturation
• layers::ops::argmax_bf16
• layers::ops::argmax_bf16_batch
• layers::ops::argmax_bf16_batch_feed
• layers::ops::argmax_bf16_batch_lp
• layers::ops::ba_gates_cross_warp_pair
• layers::ops::ba_gates_lane_kv
• layers::ops::ba_gates_log
• layers::ops::ba_gates_log_slot
• layers::ops::ba_gates_min_tokens
• layers::ops::ba_gates_output
• layers::ops::ba_gates_pick
• layers::ops::ba_gates_slot
• layers::ops::ba_gates_sm_count
• layers::ops::ba_gates_warp
• layers::ops::batched_chunk_zero_admitted
• layers::ops::batched_embed
• layers::ops::batched_embed_fp8
• layers::ops::bf16_absmax
• layers::ops::bf16_concat
• layers::ops::bf16_grouped_tc_shape_ok
• layers::ops::bf16_to_fp8
• layers::ops::check_e4m3_activation_range
• layers::ops::check_e4m3_grouped
• layers::ops::check_e4m3_range
• layers::ops::compute_gdn_gates
• layers::ops::conv1d_fwd
• layers::ops::conv1d_update
• layers::ops::conv1d_update_chunk2
• layers::ops::conv1d_update_l2norm
• layers::ops::conv1d_update_l2norm_strided
• layers::ops::conv1d_update_prefill
• layers::ops::cublas_bf16_proj_dense
• layers::ops::cublas_bf16_proj_dense_f32_out
• layers::ops::cublas_fp8_m_pad
• layers::ops::cublas_fp8_proj
• layers::ops::cublas_fp8_proj_prequant
• layers::ops::cublas_fp8_rowwise_proj
• layers::ops::cublas_scale_layout_kmajor
• layers::ops::cutlass_bf16_proj
• layers::ops::cutlass_nvfp4_proj
• layers::ops::cutlass_nvfp4_proj_from_fp8
• layers::ops::decode_w8a8_gemm
• layers::ops::decode_w8a8_quant_act
• layers::ops::decode_w8a8_selected
• layers::ops::deinterleave_qg
• layers::ops::deinterleave_qg_split
• layers::ops::deinterleave_qg_split_qnorm
• layers::ops::deinterleave_qg_split_qnorm_mrope
• layers::ops::deinterleave_qkvz
• layers::ops::dense_gemm
• layers::ops::dense_gemm_ba_gates_prefill
• layers::ops::dense_gemm_ba_gates_prefill_hopper
• layers::ops::dense_gemm_bf16_pipelined
• layers::ops::dense_gemm_m16_bf16
• layers::ops::dense_gemm_m16_bf16_n64
• layers::ops::dense_gemm_prefill
• layers::ops::dense_gemm_router
• layers::ops::dense_gemm_splitk
• layers::ops::dense_gemm_tc
• layers::ops::dense_gemm_tc_scaled_acc
• layers::ops::dense_gemv
• layers::ops::dense_gemv_ba_gates
• layers::ops::dense_gemv_batch2
• layers::ops::dense_gemv_batchm
• layers::ops::dense_gemv_batchm_split
• layers::ops::dense_gemv_fp8w
• layers::ops::dense_gemv_fp8w_batch2
• layers::ops::dense_gemv_tc::kernel_for
• layers::ops::dense_gemv_tc::launch
• layers::ops::dense_gemv_tc::mtp_tc_enabled
• layers::ops::dense_gemv_tc::mtp_tc_from
• layers::ops::dense_gemv_tc::route
• layers::ops::dense_gemv_tc::try_dense_gemv_tc
• layers::ops::dense_mm_bf16
• layers::ops::dequant_fp8_bf16_bytes
• layers::ops::dequant_fp8_bf16_into
• layers::ops::dequant_nvfp4_to_bf16
• layers::ops::dequant_q2_0_gn_to_bf16
• layers::ops::e4m3_saturations
• layers::ops::embed_from_argmax
• layers::ops::feed_resolve
• layers::ops::fill_slots_from_block_table
• layers::ops::fp4_act_scratch_bytes
• layers::ops::fp8_act_quant_hopper_enabled
• layers::ops::fp8_act_quant_hopper_reject
• layers::ops::fp8_act_scale_to_kmajor
• layers::ops::fp8_fp8_gemm_m128_mfast
• layers::ops::fp8_fp8_gemm_n128
• layers::ops::fp8_fp8_gemm_n128_m128
• layers::ops::fp8_gemm_m128_mfast
• layers::ops::fp8_gemm_n128
• layers::ops::fp8_gemm_n128_m128
• layers::ops::fp8_gemm_n128_mfast
• layers::ops::fp8_gemm_n128_row_scaled
• layers::ops::fp8_gemm_n128_row_scaled_m16
• layers::ops::fp8_gemm_t_blockscaled
• layers::ops::fp8_gemv_rowscale_batch16_rt2
• layers::ops::fp8_gemv_rowscale_batch8_rt2
• layers::ops::fp8_grouped_active_cap
• layers::ops::fp8_grouped_tc_shape_ok
• layers::ops::fp8_quant_grid
• layers::ops::fp8_quant_hopper_ctas
• layers::ops::fp8_quant_hopper_lane
• layers::ops::fp8_quant_hopper_min_m
• layers::ops::fp8_quant_hopper_span
• layers::ops::fp8_quant_log
• layers::ops::fp8_quant_log_slot
• layers::ops::fp8_quant_min_ctas
• layers::ops::fused_k_norm_rope_cache_write_bf16
• layers::ops::fused_k_norm_rope_cache_write_bf16_mrope
• layers::ops::fused_k_norm_rope_cache_write_fp8
• layers::ops::fused_k_norm_rope_cache_write_fp8_kv
• layers::ops::gated_rms_norm
• layers::ops::gated_rms_norm_prefill
• layers::ops::gated_rms_norm_strided
• layers::ops::gdn_carry_conv
• layers::ops::gdn_carry_conv_f32
• layers::ops::gdn_carry_conv_flush
• layers::ops::gdn_carry_conv_seq_elems
• layers::ops::gdn_carry_flush
• layers::ops::gdn_carry_seq_floats
• layers::ops::gdn_carry_wy
• layers::ops::gdn_conv_chain_f32
• layers::ops::gdn_conv_chain_f32_batched
• layers::ops::gdn_decode
• layers::ops::gdn_decode_chunk2
• layers::ops::gdn_decode_chunk3
• layers::ops::gdn_decode_f16_strided_norm
• layers::ops::gdn_decode_f32_conv_norm
• layers::ops::gdn_decode_f32_norm
• layers::ops::gdn_decode_f32_norm_snap
• layers::ops::gdn_decode_f32_strided
• layers::ops::gdn_decode_f32_strided_norm
• layers::ops::gdn_decode_f32_strided_norm_snap
• layers::ops::gdn_decode_strided_smem
• layers::ops::gdn_decode_wy2
• layers::ops::gdn_decode_wy3
• layers::ops::gdn_decode_wy4
• layers::ops::gdn_decode_wy4_woa
• layers::ops::gdn_decode_wyn
• layers::ops::gdn_decode_wyn_table
• layers::ops::gdn_exact_carry
• layers::ops::gdn_exact_chain
• layers::ops::gdn_exact_chain_f16
• layers::ops::gdn_init_spine_line
• layers::ops::gdn_prefill
• layers::ops::gdn_prefill_fla
• layers::ops::gdn_prefill_persistent
• layers::ops::gdn_prefill_persistent_batched
• layers::ops::gdn_prefill_persistent_smem
• layers::ops::gdn_prefill_persistent_smem_batched
• layers::ops::gdn_prefill_regresident
• layers::ops::gdn_prefill_split
• layers::ops::gdn_prefill_split4
• layers::ops::gdn_prefill_split4_batched
• layers::ops::gdn_scalar_spine
• layers::ops::gdn_scalar_spine_from
• layers::ops::gdn_tc_spine_route_line
• layers::ops::gdn_verify_fused_conv_k2
• layers::ops::gdn_verify_fused_conv_kn
• layers::ops::gdn_verify_fused_conv_kn_batched
• layers::ops::gdn_verify_fused_conv_kn_f32
• layers::ops::gdn_verify_fused_norm_k2
• layers::ops::gdn_wy4_flag_clear
• layers::ops::gdn_wy4_fold
• layers::ops::gemv_sw_from
• layers::ops::gemv_tc::narrow_gemv_max_rows
• layers::ops::gemv_tc::tc_enabled
• layers::ops::gemv_tc::tc_fixed_m
• layers::ops::gemv_tc::tc_kernel
• layers::ops::gemv_tc::tc_route
• layers::ops::gemv_tc::wide_rows_enabled
• layers::ops::grouped_gemm_mla
• layers::ops::hc_expand
• layers::ops::hc_head
• layers::ops::hc_head_lowrank
• layers::ops::hc_head_site
• layers::ops::hc_post
• layers::ops::hc_post_lowrank
• layers::ops::hc_post_site
• layers::ops::hc_pre
• layers::ops::hc_pre_lowrank
• layers::ops::hc_pre_site
• layers::ops::int8_gemm_faith2_prefill
• layers::ops::kquant_mmq_act_bytes
• layers::ops::kquant_mmq_gemm
• layers::ops::kquant_mmvq
• layers::ops::kquant_mmvq_experts
• layers::ops::kquant_mmvq_experts_w
• layers::ops::kquant_mmvq_experts_wn
• layers::ops::kquant_mmvq_groups_w
• layers::ops::kquant_mmvq_pair_w
• layers::ops::kquant_mmvq_w
• layers::ops::kquant_q8_1_rows
• layers::ops::kquant_q8_1_rows_bytes
• layers::ops::kquant_swiglu_q8_1_rows
• layers::ops::kquant_weight_bytes
• layers::ops::l2_norm
• layers::ops::log_cutlass_nvfp4_route
• layers::ops::log_gemm_shape
• layers::ops::lora_delta::apply_lora_bgmv
• layers::ops::lora_delta::apply_lora_delta
• layers::ops::lora_delta::lora_gemv_max_m
• layers::ops::lora_delta::lora_no_apply
• layers::ops::lora_delta::lora_no_ffn
• layers::ops::mamba2_ssd_bmm
• layers::ops::mamba2_ssd_cumsum
• layers::ops::mamba2_ssd_scan
• layers::ops::mamba2_ssm_prefill
• layers::ops::mamba2_ssm_prefill_persistent
• layers::ops::mla_batched_gemv
• layers::ops::mla_cache_assemble
• layers::ops::mla_cache_assemble_batched
• layers::ops::mla_fused_prefill
• layers::ops::mla_kv_assemble_batched
• layers::ops::mla_paged_decode_fp8
• layers::ops::mla_paged_decode_nvfp4
• layers::ops::mla_prefill_attention_320
• layers::ops::mla_q_final_assemble_batched
• layers::ops::mla_q_rope_extract_batched
• layers::ops::mla_q_rope_scatter
• layers::ops::mla_q_rope_writeback
• layers::ops::mla_q_rope_writeback_batched
• layers::ops::moe_act_quant_e4m3
• layers::ops::moe_batched_blend
• layers::ops::moe_bf16_grouped_gemm
• layers::ops::moe_build_tile_worklist
• layers::ops::moe_decode_atomic_c4_finalize
• layers::ops::moe_decode_atomic_c4_silu_down_accum
• layers::ops::moe_expert_down_act_bf16_grouped
• layers::ops::moe_expert_down_act_fp8_grouped
• layers::ops::moe_expert_down_act_fp8_grouped_tc_w8a8
• layers::ops::moe_expert_down_act_nvfp4_grouped
• layers::ops::moe_expert_gate_up_act_bf16_grouped
• layers::ops::moe_expert_gate_up_act_fp8_grouped
• layers::ops::moe_expert_gate_up_act_fp8_grouped_tc_w8a8
• layers::ops::moe_expert_gate_up_act_nvfp4_grouped
• layers::ops::moe_expert_gate_up_shared
• layers::ops::moe_expert_gate_up_shared_batch2
• layers::ops::moe_expert_gate_up_shared_batch2_t
• layers::ops::moe_expert_gate_up_shared_batch3
• layers::ops::moe_expert_gate_up_shared_batch3_t
• layers::ops::moe_expert_gate_up_shared_bf16
• layers::ops::moe_expert_gate_up_shared_bf16_batch2
• layers::ops::moe_expert_gate_up_shared_fp8
• layers::ops::moe_expert_gate_up_shared_fp8_batch2
• layers::ops::moe_expert_gate_up_shared_fp8_batch2_t
• layers::ops::moe_expert_gate_up_shared_fp8_batch3
• layers::ops::moe_expert_gate_up_shared_fp8_batch3_t
• layers::ops::moe_expert_gate_up_shared_fp8_t
• layers::ops::moe_expert_gate_up_shared_prefill
• layers::ops::moe_expert_gate_up_shared_t
• layers::ops::moe_expert_gemv
• layers::ops::moe_expert_gemv_gate_up
• layers::ops::moe_expert_gemv_gate_up_2x
• layers::ops::moe_expert_gemv_silu_down
• layers::ops::moe_expert_gemv_silu_down_2x
• layers::ops::moe_expert_silu_down_shared
• layers::ops::moe_expert_silu_down_shared_batch2
• layers::ops::moe_expert_silu_down_shared_batch2_t
• layers::ops::moe_expert_silu_down_shared_batch3
• layers::ops::moe_expert_silu_down_shared_batch3_t
• layers::ops::moe_expert_silu_down_shared_bf16
• layers::ops::moe_expert_silu_down_shared_bf16_batch2
• layers::ops::moe_expert_silu_down_shared_fp8
• layers::ops::moe_expert_silu_down_shared_fp8_batch2
• layers::ops::moe_expert_silu_down_shared_fp8_batch2_t
• layers::ops::moe_expert_silu_down_shared_fp8_batch3
• layers::ops::moe_expert_silu_down_shared_fp8_batch3_t
• layers::ops::moe_expert_silu_down_shared_fp8_t
• layers::ops::moe_expert_silu_down_shared_prefill
• layers::ops::moe_expert_silu_down_shared_t
• layers::ops::moe_fp8_grouped_gemm
• layers::ops::moe_fp8_grouped_gemm_ptrtable_n128
• layers::ops::moe_fp8_grouped_sort
• layers::ops::moe_gate_topk_fused
• layers::ops::moe_grouped_down_cutlass
• layers::ops::moe_grouped_gate_up_cutlass
• layers::ops::moe_hash_route
• layers::ops::moe_hash_route_batched
• layers::ops::moe_lora_grouped::gather_bgmv_grids
• layers::ops::moe_lora_grouped::gather_row_token
• layers::ops::moe_lora_grouped::grouped_down_wc
• layers::ops::moe_lora_grouped::grouped_down_windows
• layers::ops::moe_lora_grouped::moe_lora_gather_bgmv
• layers::ops::moe_lora_grouped::moe_lora_grouped_down
• layers::ops::moe_lora_grouped::pack_expert_tables
• layers::ops::moe_permute_tokens
• layers::ops::moe_router_gemm
• layers::ops::moe_router_gemm_rt
• layers::ops::moe_silu_mul
• layers::ops::moe_sort_by_expert
• layers::ops::moe_topk_sigmoid
• layers::ops::moe_topk_sigmoid_batched
• layers::ops::moe_topk_softmax
• layers::ops::moe_topk_softmax_batched
• layers::ops::moe_topk_softmax_bias
• layers::ops::moe_topk_softmax_bias_batched
• layers::ops::moe_topk_sqrtsoftplus
• layers::ops::moe_topk_sqrtsoftplus_batched
• layers::ops::moe_transpose_u8_batched
• layers::ops::moe_unpermute_blend
• layers::ops::moe_unpermute_reduce_indexed
• layers::ops::moe_w4a16_fused_gate_up_k64_m128
• layers::ops::moe_w4a16_fused_gate_up_k64_n128
• layers::ops::moe_w4a16_fused_gate_up_n128
• layers::ops::moe_w4a16_grouped_gemm
• layers::ops::moe_w4a16_grouped_gemm_ptrtable
• layers::ops::moe_w4a16_grouped_gemm_ptrtable_k64_n128
• layers::ops::moe_w4a16_grouped_gemm_ptrtable_m256
• layers::ops::moe_w4a16_grouped_gemm_ptrtable_n128
• layers::ops::moe_w4a4_grouped_gemm_relu2
• layers::ops::moe_w8a8_gateup_silu_e4m3
• layers::ops::moe_w8a8_grouped_gemm
• layers::ops::moe_w8a8_grouped_gemm_e4m3
• layers::ops::moe_w8a8_grouped_gemm_pm4
• layers::ops::moe_weighted_sum_blend
• layers::ops::moe_weighted_sum_blend_batch2
• layers::ops::moe_weighted_sum_blend_batch3
• layers::ops::moe_weighted_sum_blend_fp8_grouped
• layers::ops::moe_weighted_sum_blend_prefill
• layers::ops::moe_zero_expert_add
• layers::ops::n_tile_blocks
• layers::ops::native_q2_mmq_enabled
• layers::ops::nvfp4_grouped_tc_shape_ok
• layers::ops::nvfp4_lean_repack_shape_ok
• layers::ops::nvfp4_mmq_gemm
• layers::ops::nvfp4_mmq_gemm_tiled
• layers::ops::nvfp4_mmq_quantize_act
• layers::ops::nvfp4_mmq_repack
• layers::ops::nvfp4_mmq_smem
• layers::ops::nvfp4_mmq_weight_bytes
• layers::ops::nvfp4_pipe_gemm
• layers::ops::nvfp4_scale_bf16
• layers::ops::nvfp4_silu_mul_quant
• layers::ops::nvfp4_silu_mul_scaled
• layers::ops::nvfp4_tc_lean_repack
• layers::ops::paged_decode_attn_bf16
• layers::ops::paged_decode_attn_bf16_gqa
• layers::ops::paged_decode_attn_bf16k_turbo2v
• layers::ops::paged_decode_attn_bf16k_turbo3v
• layers::ops::paged_decode_attn_bf16k_turbo4v
• layers::ops::paged_decode_attn_fp8
• layers::ops::paged_decode_attn_fp8_gqa
• layers::ops::paged_decode_attn_fp8k_turbo2v
• layers::ops::paged_decode_attn_fp8k_turbo3v
• layers::ops::paged_decode_attn_fp8k_turbo4v
• layers::ops::paged_decode_attn_nvfp4
• layers::ops::paged_decode_attn_reduce_fp8
• layers::ops::paged_decode_attn_reduce_nvfp4
• layers::ops::paged_decode_attn_splitk_bf16
• layers::ops::paged_decode_attn_splitk_fp8
• layers::ops::paged_decode_attn_splitk_nvfp4
• layers::ops::paged_decode_attn_turbo3k_turbo8v
• layers::ops::paged_decode_attn_turbo4k_turbo3v
• layers::ops::paged_decode_attn_turbo4k_turbo8v
• layers::ops::parse_cublas_scope
• layers::ops::per_token_group_quant_fp8
• layers::ops::ple_add_highway
• layers::ops::ple_conv
• layers::ops::ple_gate
• layers::ops::predequant_nvfp4_to_fp8
• layers::ops::prefill_attention
• layers::ops::prefill_attention_512_sink
• layers::ops::prefill_attention_64
• layers::ops::prefill_attention_fp8kv
• layers::ops::prefill_attention_paged
• layers::ops::prefill_attention_paged_512
• layers::ops::prefill_attention_paged_64
• layers::ops::prefill_attention_paged_batched
• layers::ops::prefill_attention_paged_batched_64
• layers::ops::prefill_attention_paged_bf16k_turbo2v_64
• layers::ops::prefill_attention_paged_bf16k_turbo3v_64
• layers::ops::prefill_attention_paged_bf16k_turbo4v_64
• layers::ops::prefill_attention_paged_dflash
• layers::ops::prefill_attention_paged_dflash_bf16_indirect
• layers::ops::prefill_attention_paged_fp8
• layers::ops::prefill_attention_paged_fp8_64
• layers::ops::prefill_attention_paged_fp8_batched
• layers::ops::prefill_attention_paged_fp8_batched_64
• layers::ops::prefill_attention_paged_fp8_dflash
• layers::ops::prefill_attention_paged_fp8k_turbo2v_64
• layers::ops::prefill_attention_paged_fp8k_turbo3v_64
• layers::ops::prefill_attention_paged_fp8k_turbo4v_64
• layers::ops::prefill_attention_paged_nvfp4
• layers::ops::prefill_attention_paged_nvfp4_64
• layers::ops::prefill_attention_paged_nvfp4_batched
• layers::ops::prefill_attention_paged_nvfp4_batched_64
• layers::ops::prefill_attention_paged_turbo2_64
• layers::ops::prefill_attention_paged_turbo3k_turbo8v_64
• layers::ops::prefill_attention_paged_turbo4k_turbo3v_64
• layers::ops::prefill_attention_paged_turbo4k_turbo8v_64
• layers::ops::prefill_attention_paged_turbo_64
• layers::ops::prefill_batched_chunk_zero_admitted
• layers::ops::prefill_batched_first_chunk_enabled
• layers::ops::prefill_codispatch_enabled
• layers::ops::prefill_varlen_enabled
• layers::ops::q2_0_gemv
• layers::ops::q2_0_gemv_batchm
• layers::ops::q2_0_gemv_vec
• layers::ops::q2_0_gemv_vec_batchm
• layers::ops::q2_0_mmq_gemm
• layers::ops::q2_0_mmq_gemm_packed
• layers::ops::q2_0_weight_bytes
• layers::ops::q4k_mmq_gemm
• layers::ops::q4k_weight_bytes
• layers::ops::q8_1_scratch_bytes
• layers::ops::qsa_block_pool
• layers::ops::qsa_gather
• layers::ops::qsa_prefill_attn
• layers::ops::qsa_qprep
• layers::ops::qsa_qprep_rows
• layers::ops::qsa_score
• layers::ops::qsa_score_rows
• layers::ops::qsa_score_rows_tc
• layers::ops::quant_gemm
• layers::ops::quant_gemv
• layers::ops::quantize_act_q8_1
• layers::ops::quantize_bf16_to_fp8
• layers::ops::quantize_bf16_to_nvfp4
• layers::ops::quantize_weight_q4k
• layers::ops::read_expert_ptrs_u64
• layers::ops::read_expert_scales_f32
• layers::ops::requant_w_nvfp4_int8
• layers::ops::reshape_and_cache
• layers::ops::reshape_and_cache_bf16k_turbo2v
• layers::ops::reshape_and_cache_bf16k_turbo3v
• layers::ops::reshape_and_cache_bf16k_turbo4v
• layers::ops::reshape_and_cache_flash_v_only
• layers::ops::reshape_and_cache_fp8
• layers::ops::reshape_and_cache_fp8k_turbo2v
• layers::ops::reshape_and_cache_fp8k_turbo3v
• layers::ops::reshape_and_cache_fp8k_turbo4v
• layers::ops::reshape_and_cache_nvfp4
• layers::ops::reshape_and_cache_turbo3k_turbo8v
• layers::ops::reshape_and_cache_turbo4k_turbo3v
• layers::ops::reshape_and_cache_turbo4k_turbo8v
• layers::ops::residual_add
• layers::ops::residual_add_rms_norm
• layers::ops::residual_add_rms_norm_exact
• layers::ops::residual_add_rms_norm_gatef32
• layers::ops::rms_norm
• layers::ops::rms_norm_residual
• layers::ops::rms_norm_short_row_eligible
• layers::ops::rms_norm_strided
• layers::ops::rms_norm_warp_row
• layers::ops::rope
• layers::ops::rope_mrope_interleaved
• layers::ops::rope_mrope_interleaved_k_only
• layers::ops::rope_proportional
• layers::ops::rope_strided
• layers::ops::rope_yarn
• layers::ops::rope_yarn_scaled
• layers::ops::scaled_add
• layers::ops::set_prefill_codispatch_from_cli
• layers::ops::set_prefill_varlen_from_cli
• layers::ops::sigmoid_blend
• layers::ops::sigmoid_blend_device
• layers::ops::sigmoid_gate_mul
• layers::ops::sigmoid_gate_mul_batched
• layers::ops::sigmoid_gate_mul_head_broadcast
• layers::ops::silu_mul
• layers::ops::silu_mul_quant_fp8
• layers::ops::silu_mul_strided
• layers::ops::softplus_gate_mul_head_broadcast
• layers::ops::ssd_scan_fits
• layers::ops::ssd_scan_smem
• layers::ops::ssm_ba_gates_hopper_enabled
• layers::ops::ssm_ba_gates_hopper_reject
• layers::ops::ssm_h_state_f16_to_f32
• layers::ops::ssm_h_state_f32_to_f16
• layers::ops::strided_out_extent_elems
• layers::ops::target_defaults::declared
• layers::ops::target_defaults::format_levers
• layers::ops::target_defaults::resolve
• layers::ops::target_defaults::resolve_batchm_max
• layers::ops::target_defaults::resolve_max_m
• layers::ops::target_defaults::resolve_toggle
• layers::ops::target_defaults::resolved
• layers::ops::target_defaults::summary_line
• layers::ops::token_overlay::embed_overlay_routed
• layers::ops::token_overlay::embed_rowdiff
• layers::ops::token_overlay::lmhead_overlay_routed
• layers::ops::transpose_block_scale
• layers::ops::transpose_fp8
• layers::ops::transpose_u8
• layers::ops::use_gemv_sw
• layers::ops::w4a16_decode_gemv
• layers::ops::w4a16_gemm
• layers::ops::w4a16_gemm_n128
• layers::ops::w4a16_gemm_n128_ldb
• layers::ops::w4a16_gemm_n128_m128
• layers::ops::w4a16_gemm_n128_m128_bf16
• layers::ops::w4a16_gemm_n128_m128_bf16_ldb
• layers::ops::w4a16_gemm_n128_m128_v2
• layers::ops::w4a16_gemm_n128_m128_v3
• layers::ops::w4a16_gemv
• layers::ops::w4a16_gemv_batch2
• layers::ops::w4a16_gemv_batch3
• layers::ops::w4a16_gemv_batchm
• layers::ops::w4a16_gemv_dual
• layers::ops::w4a16_gemv_dual_batch2
• layers::ops::w4a16_gemv_dual_batch3
• layers::ops::w4a16_gemv_dual_sw
• layers::ops::w4a16_gemv_grid_x
• layers::ops::w4a16_gemv_qg
• layers::ops::w4a16_gemv_qg_batch2
• layers::ops::w4a16_gemv_qg_batch3
• layers::ops::w4a16_gemv_qkvz
• layers::ops::w4a16_gemv_silu_input
• layers::ops::w4a16_gemv_silu_input_sw
• layers::ops::w4a16_gemv_sw
• layers::ops::w4a16_gemv_sw_grid_x
• layers::ops::w4a16_gemv_sw_raw
• layers::ops::w4a16_t_via_fp8_ldmab
• layers::ops::w4a16_tc_rows
• layers::ops::w4a16_tc_rows_shape_ok
• layers::ops::w4a4_gemm
• layers::ops::w4a4_gemm_mfast
• layers::ops::w4a4_proj::fixed_nvfp4_proj
• layers::ops::w4a4_proj::max_rows
• layers::ops::w4a4_proj::nvfp4_proj_mx
• layers::ops::w4a4_proj::nvfp4_proj_small_m
• layers::ops::w4a4_proj::nvfp4_proj_small_m_same_input
• layers::ops::w4a4_proj::prepare
• layers::ops::w4a4_proj::ps_column_blocks
• layers::ops::w4a4_proj::ps_smem_bytes
• layers::ops::w4a4_proj::ps_stripe_chunks
• layers::ops::w4a4_proj::routes_w4a4
• layers::ops::w4a4_proj::w4a4_route
• layers::ops::w4a4_proj::weight_rows
• layers::ops::w8a16_gemm
• layers::ops::w8a16_gemm_m16
• layers::ops::w8a16_gemm_m16_n64
• layers::ops::w8a16_gemm_m16_strided
• layers::ops::w8a16_gemm_n128_m128
• layers::ops::w8a16_gemm_pipelined
• layers::ops::w8a16_gemm_pipelined_by_m
• layers::ops::w8a16_gemm_pipelined_m32
• layers::ops::w8a16_gemm_pipelined_m32_strided
• layers::ops::w8a16_gemm_pipelined_m64
• layers::ops::w8a16_gemm_pipelined_m64_strided
• layers::ops::w8a16_gemm_t
• layers::ops::w8a16_gemm_t_pipelined
• layers::ops::w8a16_gemv
• layers::ops::w8a16_gemv_batch16
• layers::ops::w8a16_gemv_batch16_ncol2
• layers::ops::w8a16_gemv_batch16_ncol2_strided
• layers::ops::w8a16_gemv_batch16_ncol4
• layers::ops::w8a16_gemv_batch16_ncol4_strided
• layers::ops::w8a16_gemv_batch16_strided
• layers::ops::w8a16_gemv_batch2
• layers::ops::w8a16_gemv_batch4
• layers::ops::w8a16_gemv_batch4_strided
• layers::ops::w8a16_gemv_dual
• layers::ops::w8a16_gemv_row_tiered
• layers::ops::w8a16_gemv_silu_input
• layers::ops::w8a16_m16_smem_bytes
• layers::ops::w8a16_pipelined_prefers_m32
• layers::ops::w8a16_pipelined_prefers_m64
• layers::ops::w8a16_tc_rows
• layers::ops::w8a16_tc_rows_launch
• layers::ops::w8a16_tc_rows_shape_ok
• layers::ops::w8a8_act_quant
• layers::ops::w8a8_act_quant_silu
• layers::ops::w8a8_decode_available
• layers::ops::w8a8_decode_proj_disabled
• layers::ops::w8a8_gemv
• layers::ops::w8a8_proj
• layers::ops::wide_prefill_kernel
• layers::ops::widen_block_scale_f32
• layers::ple::dump::tap_bf16
• layers::ple::dump::tap_f32
• layers::ple::dump::tap_highway
• layers::ple::ids::ple_ngram_ids
• layers::prefill_row_invariant
• layers::publish_row_tiers
• layers::qwen3_attention::add_attn_phase_us
• layers::qwen3_attention::add_ffn_host_us
• layers::qwen3_attention::helpers::yarn_rope_mscale
• layers::qwen3_attention::take_attn_phase_us
• layers::qwen3_attention::take_ffn_host_us
• layers::qwen3_attention::validate_required_kv_kernels
• layers::qwen3_attention::w8a8_prefill_kernels_loaded
• layers::qwen3_ssm::carry_flush_kernel
• layers::qwen3_ssm::gdn_flags::default_dflash_gamma
• layers::qwen3_ssm::gdn_flags::flags
• layers::qwen3_ssm::gdn_flags::gdn_fused_norm_enabled
• layers::qwen3_ssm::gdn_flags::gdn_woa_enabled
• layers::qwen3_ssm::gdn_flags::resolve_dflash_gamma
• layers::qwen3_ssm::gdn_flags::set_from_cli
• layers::qwen3_ssm::gdn_flags::ssm_batched_recurrent_enabled
• layers::qwen3_ssm::gdn_flags::ssm_h_dtype_bits
• layers::qwen3_ssm::gdn_flags::ssm_h_f16_pool_enabled
• layers::qwen3_ssm::gdn_flags::ssm_h_fp16_enabled
• layers::qwen3_ssm::gdn_flags::verify_exact_enabled
• layers::resolve_row_tiers
• layers::row_invariant
• layers::row_tiers
• layers::row_tiers_from
• layers::set_activation_quantization_from_cli
• layers::set_expert_quantization_from_cli
• layers::set_moe_expert_tables_from_plan
• layers::set_weight_quantization_from_cli
• layers::tgemm_kernel
• layers::tgemm_probe_ok
• layers::try_kernel
• layers::try_target_kernel
• layers::w4a16_gemv_tiers::exact_m_tiers_enabled
• layers::w4a16_gemv_tiers::select_tier
• layers::w4a16_v2_kernel
• layers::w4a16_v3_kernel
• layers::weight_quantization
• lora::adapter_id_hash
• lora::allow_partial_targets
• lora::apply_expert_lora_sorted
• lora::apply_router_lora
• lora::build_moe_row_adapter_decode
• lora::build_moe_row_adapter_host
• lora::build_overlay
• lora::build_override_set
• lora::build_seq_slot_host
• lora::clamp_trainable_to_vocab
• lora::classify_key
• lora::classify_overlay_key
• lora::ensure_decode_route_servable
• lora::expert_delta_workitems
• lora::expert_router_bytes
• lora::full_attention_layers
• lora::is_gdn_key
• lora::load_lora_adapters_generic
• lora::load_lora_adapters_multi
• lora::lora_eager_env
• lora::lora_experts_env
• lora::lora_peer_env
• lora::lora_rotate_env
• lora::max_lora_expert_rank
• lora::no_batch_verify
• lora::override_source
• lora::pack_store_into_slot
• lora::prefill_bgmv_forced
• lora::rdma_stage::build_land_targets
• lora::rdma_stage::fetch_adapter_manifest
• lora::rdma_stage::rebuild_slot_layers
• lora::rdma_stage::slot_bytes
• lora::reject_pending_overlay
• lora::resolve_moe_lora_route
• lora::routed_prefill_slot_of
• lora::router_dims
• lora::select_routed_pair
• lora::select_victim_slot
• lora::stage_overlay_raw
• lora::uniform_seq_slots
• lora::validate_peft_config
• model_type_ships_vanilla_norm_weights
• mtp_carry::carry_armed
• mtp_carry::carry_armed_with
• mtp_carry::hidden_row_offset
• mtp_carry::marconi_min_tokens
• mtp_carry::merge_interval
• mtp_carry::mtp_carry_debug
• mtp_carry::mtp_carry_drafter_enabled
• mtp_carry::plan_append
• mtp_carry::set_marconi_min_tokens
• mtp_carry::stamped_merge
• quant_format::detect_quant_format
• requires_single_chunk_prefill
• ships_vanilla_norm_weights
• speculative::draft_stop::chain_continues
• speculative::draft_stop::chain_depth
• speculative::draft_stop::draft_stop_logprob
• speculative::draft_stop::set_draft_confidence_stop
• speculative::hidden_fingerprint
• speculative::ladder::ladder_drafts_from_steps
• speculative::ladder::mtp_ladder_disabled
• speculative::ladder::mtp_ladder_drafts
• speculative::ladder::mtp_ladder_pinned
• speculative::ladder::mtp_max_seqs
• speculative::ladder::parse_mtp_k_ladder
• speculative::ladder::resolve_mtp_max_seqs
• speculative::ladder::set_mtp_k_ladder
• speculative::ladder::set_mtp_max_seqs
• speculative::mtp_accept_debug
• speculative::mtp_catchup_enabled
• speculative::mtp_ep_propose_enabled
• speculative::mtp_multi_seq_mode
• speculative::mtp_refeed_accepted_enabled
• speculative::mtp_refeed_debug
• speculative::mtp_refeed_shift
• speculative::shadow_topk
• speculative::verify_key::canonical_assignment
• speculative::verify_key::canonical_key_min_width
• speculative::verify_key::canonical_verify_key_enabled
• speculative::verify_key::verify_batch_order
• speculative::verify_key::verify_batch_permutation
• speculative::verify_key::verify_graph_key
• ssm_reserve::copy_tier
• ssm_reserve::decode_rollback_ring_slots
• ssm_reserve::decode_rollback_ring_slots_with
• ssm_reserve::fit_decode_ring_slots
• ssm_reserve::marconi_reserve_full
• ssm_reserve::marconi_snapshot_slots
• ssm_reserve::marconi_snapshot_slots_with
• ssm_reserve::mtp_pool_full_width
• ssm_reserve::mtp_state_slots
• ssm_reserve::mtp_state_slots_with
• ssm_reserve::parse_decode_ring_slots
• ssm_reserve::pool_counts
• ssm_reserve::pool_counts_tiered
• ssm_reserve::pool_counts_with
• ssm_reserve::prefix_caching_active
• ssm_reserve::published_decode_ring_slots
• ssm_reserve::recurrent_units
• ssm_reserve::set_copy_tier
• ssm_reserve::set_decode_ring_slots
• ssm_reserve::set_ssm_rollback_mode
• ssm_reserve::ssm_h_prefill_stage_bytes
• ssm_reserve::ssm_h_stored_bytes
• ssm_reserve::ssm_replay_ring_bytes
• ssm_reserve::ssm_replay_row_bytes
• ssm_reserve::ssm_rollback_mode
• ssm_reserve::state_dims
• ssm_reserve::tier_h
• ssm_reserve::tier_rows
• ssm_reserve::verify_slot_drafts
• ssm_reserve::verify_slot_drafts_with
• ssm_reserve::verify_slot_h_intermediates
• ssm_reserve::watchdogs_disabled_from_value
• video_decode_ffmpeg::decode_frames
• video_decode_ffmpeg::probe
• video_preprocess::decode_frames
• video_preprocess::preprocess_video
• video_preprocess::sample_indices
• vision_preprocess::preprocess_image
• vision_preprocess::preprocess_image_with_max_pixels
• vision_preprocess_glm::block_major_patch_index
• vision_preprocess_glm::pixel_bounds
• vision_preprocess_glm::resize_and_pad
• vision_preprocess_glm::smart_resize
• weight_map::fp8_dequant::dequant_fp8_to_bf16_into
• weight_map::fp8_lut::dequant_nvfp4_to_bf16
• weight_map::fp8_lut::gpu_concat_rows
• weight_map::fp8_lut::interleave_ba
• weight_map::fp8_lut::load_dense_ffn
• weight_map::load_fp8_weight
• weight_map::loaders_fp8::load_attention
• weight_map::loaders_fp8::load_fp8_block_scaled_as_fp8weight
• weight_map::loaders_fp8::load_moe
• weight_map::loaders_fp8::load_moe_skip_experts
• weight_map::loaders_fp8::load_ssm
• weight_map::loaders_fp8::quantize_to_nvfp4
• weight_map::loaders_moe::load_moe_mistral
• weight_map::loaders_moe::load_mtp
• weight_map::model_a::dense
• weight_map::model_a::dense_auto_fp8_or_bf16
• weight_map::model_a::dense_f32_as_bf16
• weight_map::model_a::dense_f32_safe
• weight_map::model_a::dense_keep_f32
• weight_map::model_a::load_checkpoint_kv_scales
• weight_map::model_a::load_kv_scales
• weight_map::model_a::quantized
• weight_map::model_a::quantized_mxfp4_e8m0
• weight_map::model_a::quantized_mxfp4_e8m0_pair
• weight_map::model_a::scalar_bf16_or_f32
• weight_map::nemotron::load_nemotron_attention
• weight_map::nemotron::load_nemotron_moe
• weight_map::nemotron::load_nemotron_ssm
• weight_map::nvfp4_detect::config_declared_variant
• weight_map::nvfp4_detect::detect_nvfp4_variant
• weight_map::nvfp4_detect::quantized_any
• weight_map::nvfp4_detect::quantized_auto
• weight_map::nvfp4_detect::quantized_from_fp8
• weight_map::quant_helpers::dense_auto
• weight_map::quant_helpers::dequant_fp8_blockscaled_to_bf16
• weight_map::quant_helpers::quantized_v2
• weight_map::quantize_to_fp8
• weight_map::quantize_to_fp8_blockscaled
• weight_map::ssm_qwen35::load_moe_no_shared
• weight_map::ssm_qwen35::load_moe_qwen35
• weight_map::ssm_qwen35::load_moe_qwen35_fp8_experts
• weight_map::ssm_qwen35::load_ssm_qwen35
• weight_map::ssm_qwen35_more::load_moe_gemma4
• weight_map::ssm_qwen35_more::load_moe_minimax

Type Aliases

• layers::ops::ContiguousBatchGemv
• layers::ops::ContiguousM16Gemm
• layers::ops::DenseM16Bf16Gemm

Statics

• layers::qwen3_attention::ATTN_PHASE_US
• layers::qwen3_attention::FFN_HOST_US

Constants

• circuit_exec::policy::MULTI_SEQ_ENV_SWITCHES
• circuit_exec::program::MAX_VERIFY_STEPS
• circuit_exec::routes::KNOWN
• circuit_exec::sources::BLOCKS
• circuit_exec::sources::CHECKPOINTS
• circuit_exec::sources::CIRCUITS
• circuit_exec::sources::FUSIONS
• circuit_exec::sources::INSTANCES
• circuit_exec::sources::PRECISION
• circuit_exec::spec_key::KEY_MODES
• circuit_exec::verify_batch::CACHE_CAP
• drafter_context::DISABLE_ENV
• drafter_context::PREFILL_ONLY_ENV
• layer::MTP_CATCHUP_MAX
• layer::VERIFY_WY_LAYER_STRIDE_BYTES
• layer::VERIFY_WY_TABLES_PER_LAYER
• layer::VERIFY_WY_TABLE_SEQS
• layer::VERIFY_WY_TABLE_STRIDE_BYTES
• layers::dense_ffn::m16_tc::oracle::F32_UNIT_ROUNDOFF
• layers::dense_ffn::m16_tc::oracle::M16_TC_ACC_FLOOR_MARGIN
• layers::dense_ffn::m16_tc::oracle::M16_TC_MAX_ULP
• layers::dense_ffn::m16_tc::oracle::M16_TC_SIGN_FLIP_BAND
• layers::moe::NVFP4_GROUPED_DECODE_MAX_ROWS
• layers::moe::NVFP4_GROUPED_DECODE_TC_MAX_ROWS
• layers::mtp_meta::MTP_META_OFFSET
• layers::ops::BA_GATES_BLOCK
• layers::ops::BA_GATES_FALLBACK_SM_COUNT
• layers::ops::BA_GATES_GROUPS
• layers::ops::BA_GATES_LANES
• layers::ops::BA_GATES_LOG_SLOTS
• layers::ops::BA_GATES_OUTS
• layers::ops::BA_GATES_REJECTS
• layers::ops::BA_GATES_TOO_FEW_TOKENS
• layers::ops::BA_GATES_WARPS
• layers::ops::BF16_GROUPED_DOWN_TC
• layers::ops::BF16_GROUPED_GATE_UP_TC
• layers::ops::DECODE_W8A8_ROWS
• layers::ops::DENSE_GEMM_M16_BF16_K_STEP
• layers::ops::DENSE_GEMM_M16_BF16_MAX_M
• layers::ops::DENSE_GEMM_M16_BF16_N_TILE
• layers::ops::DENSE_GEMM_M16_BF16_N_TILE_WIDE
• layers::ops::DENSE_GEMV_BATCHM_DECODE_MAX_M
• layers::ops::DENSE_GEMV_BATCHM_MAX_M
• layers::ops::E4M3_MAX
• layers::ops::FEED_HOST_BIT
• layers::ops::FP8_GROUPED_DOWN_COLS_PER_CTA
• layers::ops::FP8_GROUPED_DOWN_ROWS_PER_PASS
• layers::ops::FP8_GROUPED_DOWN_SCALAR
• layers::ops::FP8_GROUPED_DOWN_TC
• layers::ops::FP8_GROUPED_DOWN_TC_W8A8
• layers::ops::FP8_GROUPED_GATE_UP_COLS_PER_CTA
• layers::ops::FP8_GROUPED_GATE_UP_ROWS_PER_PASS
• layers::ops::FP8_GROUPED_GATE_UP_SCALAR
• layers::ops::FP8_GROUPED_GATE_UP_TC
• layers::ops::FP8_GROUPED_GATE_UP_TC_W8A8
• layers::ops::FP8_GROUPED_SORT_MAX_EXPERTS
• layers::ops::FP8_GROUPED_TC_DOWN_COLS_PER_CTA
• layers::ops::FP8_GROUPED_TC_GATE_UP_COLS_PER_CTA
• layers::ops::FP8_GROUPED_TC_ROWS_PER_PASS
• layers::ops::FP8_QUANT_ENTRY
• layers::ops::FP8_QUANT_HOPPER_ENTRY
• layers::ops::FP8_QUANT_HOPPER_GROUPS_PER_CTA
• layers::ops::FP8_QUANT_HOPPER_MODULE
• layers::ops::FP8_QUANT_LOG_SLOTS
• layers::ops::FP8_QUANT_MIN_CTAS_PER_SM
• layers::ops::FP8_QUANT_MODULE
• layers::ops::FP8_QUANT_REJECTS
• layers::ops::FP8_QUANT_TOO_FEW_CTAS
• layers::ops::GDN_CARRY_CAP
• layers::ops::GDN_CARRY_EAGER_MAX_PENDING
• layers::ops::GDN_CARRY_LAZY_MIN_SEQS
• layers::ops::GDN_DECODE_STRIDED_SMEM_CAP_BYTES
• layers::ops::GDN_DECODE_STRIDED_SMEM_CAP_MIN_BATCH
• layers::ops::GDN_SCALAR_SPINE_PIPE
• layers::ops::GDN_SCALAR_SPINE_VFUSED
• layers::ops::GDN_SCALAR_SPINE_VTILE
• layers::ops::GDN_TC_SPINE_ENTRY
• layers::ops::GDN_TC_SPINE_MODULE
• layers::ops::KQUANT_MODULE
• layers::ops::MAX_DYNAMIC_SMEM
• layers::ops::MIN_CTAS_PER_SM
• layers::ops::MOE_E4M3_DN
• layers::ops::MOE_E4M3_GU
• layers::ops::MOE_ROUTED_DOWN_KNOWN
• layers::ops::MOE_ROUTER_GEMM_COLS
• layers::ops::MOE_ROUTER_GEMM_MAX_PREFILL_ROWS
• layers::ops::MOE_ROUTER_GEMM_ROWS
• layers::ops::MOE_ROUTER_RT_MIN_ROWS
• layers::ops::MOE_TOPK_SIGMOID_MAX_EXPERTS
• layers::ops::MOE_TOPK_SIGMOID_MAX_TOP_K
• layers::ops::NEMOTRON_SHARED_EXPERT_E4M3
• layers::ops::NVFP4_BLOCK_BYTES
• layers::ops::NVFP4_GROUPED_DOWN_COLS_PER_CTA
• layers::ops::NVFP4_GROUPED_DOWN_ROWS_PER_PASS
• layers::ops::NVFP4_GROUPED_DOWN_SCALAR
• layers::ops::NVFP4_GROUPED_DOWN_TC
• layers::ops::NVFP4_GROUPED_GATE_UP_COLS_PER_CTA
• layers::ops::NVFP4_GROUPED_GATE_UP_ROWS_PER_PASS
• layers::ops::NVFP4_GROUPED_GATE_UP_SCALAR
• layers::ops::NVFP4_GROUPED_GATE_UP_TC
• layers::ops::NVFP4_GROUPED_TC_ROWS_PER_PASS
• layers::ops::NVFP4_LEAN_REPACK_SMEM_PER_K
• layers::ops::NVFP4_MMQ_SMEM
• layers::ops::NVFP4_PIPE_SMEM
• layers::ops::Q2K_BLOCK_BYTES
• layers::ops::Q2K_MMQ_SMEM
• layers::ops::Q2_0_BLOCK_BYTES
• layers::ops::Q2_BATCHM_MAX_M
• layers::ops::Q3K_BLOCK_BYTES
• layers::ops::Q3K_MMQ_SMEM
• layers::ops::Q4K_BLOCK_BYTES
• layers::ops::Q4K_MMQ_SMEM
• layers::ops::Q6K_BLOCK_BYTES
• layers::ops::Q8_1_BLOCK_BYTES
• layers::ops::QK2_0
• layers::ops::QK_K
• layers::ops::QK_NVFP4
• layers::ops::SSD_L
• layers::ops::SSD_PT
• layers::ops::W4A16_GEMV_OUTS_PER_BLOCK
• layers::ops::W4A16_GEMV_SW_OUTS_PER_BLOCK
• layers::ops::W4A16_TC_ROWS_COLS
• layers::ops::W4A16_TC_ROWS_MAX_M
• layers::ops::W4A16_TC_ROWS_MODULE
• layers::ops::W4A16_VIA_FP8_MIN_M
• layers::ops::W8A16_GEMM_M16_N_TILE
• layers::ops::W8A16_GEMM_M16_N_TILE_WIDE
• layers::ops::W8A16_M16_MAX_K
• layers::ops::W8A16_M16_MAX_ROWS
• layers::ops::W8A16_M32_K_BLOCK
• layers::ops::W8A16_M32_TILE_ROWS
• layers::ops::W8A16_M64_TILE_ROWS
• layers::ops::W8A16_TC_ROWS_COLS
• layers::ops::W8A16_TC_ROWS_MAX_M
• layers::ops::W8A8_K_UNIT
• layers::ops::W8A8_LAUNCH_ROWS
• layers::ops::W8A8_MAX_ROWS
• layers::ops::dense_gemv_tc::BLOCK
• layers::ops::dense_gemv_tc::K_STEP
• layers::ops::dense_gemv_tc::MIN_M
• layers::ops::dense_gemv_tc::ROWS_PER_CTA
• layers::ops::dense_gemv_tc::TC16_MAX_M
• layers::ops::dense_gemv_tc::TC32_MAX_M
• layers::ops::dense_gemv_tc::TC8_MAX_M
• layers::ops::gemv_tc::NARROW_MAX_ROWS
• layers::ops::gemv_tc::TC16_COLS_PER_CTA
• layers::ops::gemv_tc::TC16_MAX_M
• layers::ops::gemv_tc::TC8_COLS_PER_CTA
• layers::ops::gemv_tc::TC8_MAX_M
• layers::ops::gemv_tc::TC_BLOCK
• layers::ops::gemv_tc::WIDE_MAX_ROWS
• layers::ops::moe_lora_grouped::MLG_M_TILE
• layers::ops::target_defaults::BASELINE_BATCHM_MAX
• layers::ops::w4a4_proj::PS_MIN_TILES_PER_SM
• layers::ops::w4a4_proj::PS_SMEM_MAX
• layers::ops::w4a4_proj::W4A4_MAX_K
• layers::ops::w4a4_proj::W4A4_MAX_M
• layers::ops::w4a4_proj::W4A4_WIDE_MAX_M
• layers::qwen3_attention::W8A8_PREFILL_KERNELS
• layers::qwen3_ssm::gdn_flags::MAX_F16_TWIN_DFLASH_GAMMA
• layers::qwen3_ssm::gdn_flags::MAX_F16_TWIN_K
• layers::qwen3_ssm::gdn_flags::UNRESOLVED_DFLASH_GAMMA
• layers::vision_encoder::IMAGE_PAD_TOKEN_ID
• layers::vision_encoder::VIDEO_PAD_TOKEN_ID
• layers::w4a16_gemv_tiers::W4A16_BATCHM_WIDTHS
• lora::ROWDIFF_THRESH
• mtp_carry::DEFAULT_MARCONI_MIN_TOKENS
• speculative::EP_CMD_MTP_PROPOSE
• speculative::ladder::DEFAULT_MTP_MAX_SEQS
• speculative::ladder::MAX_MTP_MAX_SEQS
• speculative::tree_shape::MAX_NODES
• speculative::tree_shape::MAX_RANK
• speculative::verify_key::CANONICAL_KEY_MIN_WIDTH
• ssm_reserve::DECODE_RING_FIT_LADDER
• ssm_reserve::MTP_STATE_SLOTS_FLOOR
• video_preprocess::DEFAULT_FPS
• video_preprocess::DEFAULT_MAX_FRAMES
• video_preprocess::DEFAULT_MIN_FRAMES