List of all items

Structs

• engine::GenerateResult
• factory::DflashBuildArgs
• factory::LoraBuildArgs
• 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::GdnPrefillBuffers
• layer::MidchunkCapture
• layer::SsmLayerState
• layers::deepseek_v4_mtp::DeepseekV4MtpHead
• layers::deepseek_v4_mtp::DeepseekV4MtpProposerState
• layers::dense_ffn::DenseFfnLayer
• layers::dense_ffn::DenseFfnWeights
• layers::dense_ffn::DenseFfnWeightsBf16
• layers::dense_ffn::DenseFfnWeightsFp8
• layers::dense_ffn::DenseFfnWeightsQ2
• layers::dflash_head::BlockDiffusionDraftHead
• layers::dflash_head::DflashKernels
• layers::dflash_head::DflashLayer
• layers::dflash_head::DflashProposerState
• layers::dflash_head::DflashScratch
• layers::ep_dispatch::EpRoutingTable
• layers::fp8_calibration::Fp8KvCalibration
• layers::glm5next_dsa::Glm5NextDsaConfig
• layers::glm5next_dsa::Glm5NextDsaKernels
• layers::glm5next_dsa::attend::DsaDecodeInputs
• layers::glm5next_dsa::attend::DsaDecodePaging
• layers::glm5next_dsa::attend::Glm5NextDsaDecodeKernel
• layers::glm5next_dsa::binding::DsaBindReport
• layers::glm5next_dsa::binding::DsaSpec
• layers::glm5next_dsa::layer::Glm5NextDsaLayer
• layers::glm5next_dsa::layer::Glm5NextDsaLayerKernels
• layers::glm5next_dsa::layer::Glm5NextDsaWeights
• layers::glm5next_dsa::layer::Glm5NextDsaWorkspace
• layers::glm5next_dsa::select::DsaSelectGeometry
• layers::glm5next_dsa::select::DsaSelectInputs
• layers::glm5next_dsa::select::DsaSelectScratch
• layers::glm5next_dsa::state::Glm5NextDsaState
• layers::glm5next_dsa::tp::DsaTensorPlan
• layers::glm5next_dsa::tp::DsaTpPlan
• layers::glm5next_dsa_ref::DsaDims
• layers::glm5next_dsa_ref::Pools
• layers::glm5next_kda::Glm5NextKdaConfig
• layers::glm5next_kda::Glm5NextKdaKernels
• layers::glm5next_kda::Glm5NextKdaLayer
• layers::glm5next_kda::Glm5NextKdaWeights
• layers::glm5next_kda::Glm5NextKdaWorkspace
• layers::glm5next_kda::KdaSeqState
• layers::glm5next_kda::binding::KdaBindReport
• layers::glm5next_kda::binding::RawTensor
• layers::glm5next_kda::binding::TensorSpec
• layers::glm5next_kda::tp::KdaTensorPlan
• layers::glm5next_kda::tp::KdaTpPlan
• layers::glm5next_kda::tp_bind::KdaShardedSource
• layers::glm5next_kda_ref::KdaDims
• layers::glm5next_kda_ref::KdaWeights
• layers::glm5next_layer::Glm5NextLayer
• layers::glm5next_layer::Glm5NextMhc
• layers::glm5next_mlp::Glm5NextMlpConfig
• layers::glm5next_mlp::Glm5NextMlpKernels
• layers::glm5next_mlp::forward::Glm5NextMlpWorkspace
• layers::glm5next_mlp::weights::Glm5NextDenseMlpWeights
• layers::glm5next_mlp::weights::Glm5NextExpertPtrTable
• layers::glm5next_mlp::weights::Glm5NextExpertWeights
• layers::glm5next_mlp::weights::Glm5NextMoePtrTables
• layers::glm5next_mlp::weights::Glm5NextMoeWeights
• layers::glm5next_mlp::weights::Nvfp4Proj
• layers::glm5next_mtp_head::Glm5NextMtpHead
• layers::glm5next_mtp_head::Glm5NextMtpProposerState
• layers::glm5next_skeleton::Glm5NextTextSkeleton
• layers::glm5next_skeleton::SkeletonLayer
• layers::glm5next_skeleton::StateBudget
• layers::glm5next_skeleton::StructuralAccounting
• layers::moe::MoeLayer
• layers::mtp_head::MtpHead
• layers::mtp_head::MtpProposerState
• layers::mtp_multi::MultiModuleMtpHead
• layers::mtp_multi::MultiModuleMtpState
• layers::nemotron_mamba2::NemotronMamba2Layer
• layers::nemotron_moe::NemotronMoeLayer
• layers::ngram_embed::NgramDims
• layers::ngram_embed::NgramEmbedding
• layers::ops::DenseMmKernels
• layers::ops::DerivedWeights
• layers::ops::GemmDispatch
• layers::ops::Glm5NextMhcKernels
• layers::ops::Glm5NextMhcSiteWeights
• layers::ops::ModelLevers
• layers::ops::MoeCutlassDownHostTables
• layers::ops::MoeCutlassHostTables
• 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::token_overlay::OverlayKernels
• 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::HcHeadWeights
• layers::qwen3_attention::HcLowRank
• layers::qwen3_attention::HcSiteWeights
• layers::qwen3_attention::HcWeights
• layers::qwen3_attention::MlaWeights
• layers::qwen3_attention::Qwen3AttentionLayer
• layers::qwen3_attention::innerq_driver::InnerQDriver
• 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
• mistral_loader::MistralWeightLoader
• model::TransformerModel
• model::nllb::NllbGpuModel
• model::nllb::NllbLang
• precision_schedule::PrecisionSchedule
• quant_format::CompressedTensorsFormat
• quant_format::Fp8BlockScaledFormat
• quant_format::ModeloptFormat
• seq_state_reserve::PerSequenceState
• speculative::tree_shape::HedgeNode
• speculative::tree_shape::TreeDraft
• speculative::tree_shape::TreeRow
• speculative::tree_shape::TreeShape
• ssm_reserve::DecodeRingDecision
• ssm_reserve::MarconiSlotDecision
• tp_shard::TpAttentionDims
• tp_shard::TpGdnDims
• tp_shard::TpMoeDims
• traits::BeamReq
• traits::ChunkedPrefillPageMetadata
• traits::EpCommandFailed
• traits::MixedBatchResult
• traits::MixedForwardResult
• traits::PrefillSlice
• traits::PromptTokenLogprob
• traits::SequenceState
• video_decode_ffmpeg::FfmpegPolicy
• video_preprocess::PreprocessedVideo
• vision_item::VisionItem
• weight_loader::DeepSeekV4WeightLoader
• weight_loader::Gemma4WeightLoader
• weight_loader::LagunaWeightLoader
• weight_loader::LongcatWeightLoader
• weight_loader::MinimaxM2WeightLoader
• weight_loader::NemotronHWeightLoader
• weight_loader::NllbWeightLoader
• weight_loader::Qwen35DenseWeightLoader
• weight_loader::Qwen35WeightLoader
• weight_loader::Qwen3VLWeightLoader
• weight_loader::Qwen3WeightLoader
• weight_loader::Qwen4ExpWeightLoader
• weight_loader::Step3p7WeightLoader
• weight_loader::dflash_loader::DflashConfig
• weight_loader::dflash_loader::DflashLayerWeights
• weight_loader::dflash_loader::DflashRopeScaling
• weight_loader::dflash_loader::DflashSubConfig
• weight_loader::dflash_loader::DflashWeights
• weight_loader::glm5_next::Accounting
• weight_loader::glm5_next_load::Glm5NextWeightLoader
• weight_map::AttentionWeights
• weight_map::DenseExpertWeight
• weight_map::DenseWeight
• weight_map::ExpertWeight
• weight_map::Fp8DenseWeight
• weight_map::Fp8ExpertWeight
• weight_map::Fp8Weight
• weight_map::Fp8WeightTransposed
• weight_map::ModelWeights
• weight_map::MoeWeights
• weight_map::MtpWeights
• weight_map::NemotronExpertWeight
• weight_map::NemotronMoeWeights
• weight_map::NemotronSsmWeights
• weight_map::PackedQ2Weight
• weight_map::QuantExpertWeight
• weight_map::QuantizedWeight
• weight_map::SsmWeights
• weight_map::SsmWeightsQwen35

Enums

• forward::qwen3_5::MetalKvDtype
• layer::MoeLoraRoute
• layers::FfnComponent
• layers::dense_ffn::FfnActivation
• layers::dflash_head::DflashQuantization
• layers::glm5next_dsa::select::DsaSelectLaunch
• layers::glm5next_dsa::tp::DsaShard
• layers::glm5next_kda::binding::AttnBlockKind
• layers::glm5next_kda::binding::KdaDtype
• layers::glm5next_kda::tp::KdaShard
• layers::glm5next_layer::Glm5NextMixer
• layers::glm5next_layer::Glm5NextMlpSite
• layers::glm5next_mlp::Glm5NextMlpKind
• layers::glm5next_skeleton::FinalStep
• layers::glm5next_skeleton::Mixer
• layers::glm5next_skeleton::Mlp
• layers::glm5next_skeleton::ResidualStep
• layers::glm5next_skeleton::Site
• layers::glm5next_skeleton::StateKind
• layers::mtp_head::MtpQuantization
• layers::ngram_embed::NgramTable
• layers::ops::Derivation
• layers::ops::HcVariant
• lora::AdapterAb
• lora::ExpertProj
• lora::LoraModule
• lora::LoraTarget
• lora::OverlayModule
• lora::OverlayTensorKind
• lora::VictimError
• mtp_layout::MtpLayout
• precision_schedule::Dtype
• precision_schedule::Role
• ssm_reserve::SsmRollbackMode
• tp_shard::TpShardKind
• video_decode_ffmpeg::Availability
• weight_loader::QuantFormat
• weight_loader::WeightFormat
• weight_loader::glm5_next::TensorRole
• weight_map::LayerWeights
• weight_map::NemotronSsmQuant
• weight_map::Nvfp4Variant
• weight_map::QuantWeight
• weight_map::WeightQuantFormat

Traits

• forward::quant_weights::QuantWeights
• layer::LayerState
• layer::TransformerLayer
• layers::glm5next_kda::binding::KdaTensorSource
• quant_format::QuantFormat
• speculative::DraftProposer
• speculative::ProposerState
• traits::Model
• weight_loader::ModelWeightLoader

Functions

• engine::generate
• engine::generate_speculative
• engine::generate_streaming
• factory::build_model
• factory::loader_for_config
• forward::qwen3_5::forward_full_attention
• forward::qwen3_5::forward_linear_attention
• layers::dflash_head::dflash_ctx_cap
• layers::ep_dispatch::build_ep_routing_table
• layers::fp8_calibration::dtype_runs_online_fp8_kv_calibration
• layers::fp8_calibration::graphs_ready_after_fp8_kv_cal
• layers::glm5next_dsa::attend::decode_attention
• layers::glm5next_dsa::binding::dsa_tensor_specs
• layers::glm5next_dsa::binding::verify_dsa_block
• layers::glm5next_dsa::build::absorb_o
• layers::glm5next_dsa::build::absorb_q
• layers::glm5next_dsa::build::build_dsa_weights
• layers::glm5next_dsa::build::shard_host
• layers::glm5next_dsa::select::contiguous_pool_count
• layers::glm5next_dsa::select::select_tokens
• layers::glm5next_dsa::select::topk_smem_for_tile
• layers::glm5next_dsa::select::topk_tile
• layers::glm5next_dsa::state::dsa_capacity
• layers::glm5next_dsa::state::indexer_state_bytes
• layers::glm5next_dsa::state::max_dsa_context
• layers::glm5next_dsa_ref::expand_kv
• layers::glm5next_dsa_ref::expand_selection
• layers::glm5next_dsa_ref::index_scores
• layers::glm5next_dsa_ref::kept_pools
• layers::glm5next_dsa_ref::layer_norm
• layers::glm5next_dsa_ref::linear
• layers::glm5next_dsa_ref::mla_masked_attention
• layers::glm5next_dsa_ref::pool_states
• layers::glm5next_dsa_ref::rms_norm
• layers::glm5next_dsa_ref::sigmoid_f32
• layers::glm5next_dsa_ref::topk_pools
• layers::glm5next_dsa_ref::topk_to_mask
• layers::glm5next_dsa_ref::visible
• layers::glm5next_kda::binding::bind_kda_weights
• layers::glm5next_kda::binding::classify_attn_block
• layers::glm5next_kda::tp_bind::shard_bytes
• layers::glm5next_kda_ref::bounded_gate
• layers::glm5next_kda_ref::kda_chunked
• layers::glm5next_kda_ref::kda_chunked_prenorm
• layers::glm5next_kda_ref::kda_recurrent
• layers::glm5next_kda_ref::kda_recurrent_prenorm
• layers::glm5next_kda_ref::kda_reference_layer
• layers::glm5next_kda_ref::l2norm_rows
• layers::glm5next_kda_ref::linear
• layers::glm5next_kda_ref::rms_norm_gated
• layers::glm5next_kda_ref::unbounded_gdn_gate
• layers::glm5next_layer::profile::end
• layers::glm5next_layer::profile::end_nosync
• layers::glm5next_layer::profile::end_us
• layers::glm5next_layer::profile::full
• layers::glm5next_layer::profile::on
• layers::glm5next_layer::profile::probe_buf
• layers::glm5next_layer::profile::start
• layers::glm5next_layer::profile::start_hot
• layers::glm5next_layer::profile::stash_route
• layers::glm5next_layer::profile::step
• layers::glm5next_layer::profile::trace_bar
• layers::glm5next_layer::profile::trace_on
• layers::glm5next_layer::state::alloc_kda_ssm_state
• layers::glm5next_mlp::build::build_dense_mlp
• layers::glm5next_mlp::build::build_moe
• layers::glm5next_mlp::forward::forward_dense
• layers::glm5next_mlp::forward::forward_moe
• layers::k64_kernel
• layers::k64_n64_kernel
• layers::k64_n64_wins
• 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::ngram_embed::ngram_ids
• layers::ops::argmax_bf16
• layers::ops::argmax_bf16_batch
• layers::ops::argmax_bf16_batch_lp
• layers::ops::batched_embed
• layers::ops::batched_embed_fp8
• layers::ops::bf16_absmax
• layers::ops::bf16_concat
• layers::ops::bf16_to_fp8
• 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
• layers::ops::cublas_bf16_proj_dense
• layers::ops::cublas_fp8_proj
• layers::ops::cublas_fp8_rowwise_proj
• layers::ops::cutlass_bf16_proj
• layers::ops::cutlass_nvfp4_proj
• layers::ops::cutlass_nvfp4_proj_from_fp8
• 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_bf16_pipelined
• 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_fp8w
• layers::ops::dense_gemv_fp8w_batch2
• layers::ops::dense_mm_bf16
• layers::ops::dequant_nvfp4_to_bf16
• layers::ops::dequant_q2_0_gn_to_bf16
• layers::ops::embed_from_argmax
• layers::ops::fill_slots_from_block_table
• layers::ops::fp4_act_scratch_bytes
• 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::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::gated_rms_norm
• layers::ops::gated_rms_norm_prefill
• 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_wy2
• layers::ops::gdn_decode_wy3
• layers::ops::gdn_decode_wy4
• layers::ops::gdn_decode_wyn
• layers::ops::gdn_flashinfer::available
• layers::ops::gdn_flashinfer::flashinfer_gdn_prefill
• 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_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::gemv_sw_from
• layers::ops::glm_hc_expand
• layers::ops::glm_hc_post
• layers::ops::glm_hc_pre
• layers::ops::grouped_gemm_mla
• layers::ops::hc_expand
• layers::ops::hc_head
• layers::ops::hc_head_lowrank
• layers::ops::hc_head_mean
• 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::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::mix_hc
• 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_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_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_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_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_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_grouped_gemm
• 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_prefill
• layers::ops::moe_zero_expert_add
• layers::ops::native_q2_mmq_enabled
• 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_scale_bf16
• layers::ops::nvfp4_silu_mul_quant
• layers::ops::nvfp4_silu_mul_scaled
• layers::ops::paged_decode_attn_bf16
• 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_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_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::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_first_chunk_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_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_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::softplus_gate_mul_head_broadcast
• layers::ops::ssd_scan_fits
• layers::ops::ssd_scan_smem
• layers::ops::ssm_h_state_f16_to_f32
• layers::ops::ssm_h_state_f32_to_f16
• 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::w4a4_gemm
• layers::ops::w4a4_gemm_mfast
• layers::ops::w8a16_gemm
• layers::ops::w8a16_gemm_n128_m128
• layers::ops::w8a16_gemm_pipelined
• layers::ops::w8a16_gemm_t
• layers::ops::w8a16_gemm_t_pipelined
• layers::ops::w8a16_gemv
• layers::ops::w8a16_gemv_batch2
• layers::ops::w8a16_gemv_batch4
• layers::ops::w8a16_gemv_dual
• layers::ops::w8a16_gemv_silu_input
• 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::qwen3_attention::add_attn_phase_us
• layers::qwen3_attention::add_ffn_host_us
• layers::qwen3_attention::take_attn_phase_us
• layers::qwen3_attention::take_ffn_host_us
• layers::qwen3_attention::validate_required_kv_kernels
• 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::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::tgemm_kernel
• layers::try_kernel
• layers::w4a16_gemv_tiers::exact_m_tiers_enabled
• layers::w4a16_gemv_tiers::select_tier
• layers::w4a16_v2_kernel
• layers::w4a16_v3_kernel
• 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::validate_peft_config
• model_type_ships_vanilla_norm_weights
• mtp_layout::detect
• mtp_layout::detect_in_store
• preflight::preflight
• quant_format::detect_quant_format
• requires_single_chunk_prefill
• seq_state_reserve::per_sequence_state_bytes
• ships_vanilla_norm_weights
• speculative::draft_conf_tau
• speculative::hidden_fingerprint
• speculative::ladder::mtp_ladder_disabled
• speculative::ladder::mtp_ladder_drafts
• speculative::ladder::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::decode_rollback_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::prefix_caching_active
• ssm_reserve::set_ssm_rollback_mode
• ssm_reserve::ssm_h_prefill_stage_bytes
• ssm_reserve::ssm_h_stored_bytes
• ssm_reserve::ssm_pool_reserve_bytes
• ssm_reserve::ssm_replay_ring_bytes
• ssm_reserve::ssm_replay_row_bytes
• ssm_reserve::ssm_rollback_mode
• ssm_reserve::verify_slot_drafts
• ssm_reserve::verify_slot_drafts_with
• ssm_reserve::verify_slot_h_intermediates
• tp_shard::load_qk_norms_tp
• tp_shard::load_qkvo_tp
• tp_shard::shard_dense_1d_bf16
• tp_shard::shard_dense_bf16
• tp_shard::shard_dense_weight
• tp_shard::shard_fp8_block_scaled
• tp_shard::shard_gdn_ba_rows
• tp_shard::shard_gdn_conv_rows
• tp_shard::shard_gdn_out_proj_row_parallel
• tp_shard::shard_gdn_qkv_rows
• tp_shard::shard_gdn_qkvz_rows
• tp_shard::shard_gdn_value_vector
• tp_shard::shard_quantized_nvfp4
• traits::bf16_to_f32
• traits::extract_bf16
• traits::logprob_of
• traits::padded_batch_n
• 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
• weight_loader::dflash_loader::load_dflash_weights
• weight_loader::dflash_loader::parse_dflash_config
• weight_loader::dflash_loader::store_has_dflash_weights
• weight_loader::glm5_next::account
• weight_loader::glm5_next::classify
• weight_loader::glm5_next::is_mtp_only_name
• weight_loader::glm5_next::layer_index
• weight_map::detect_nvfp4_variant
• weight_map::load_fp8_block_scaled_as_fp8weight
• weight_map::load_fp8_weight
• weight_map::quantize_to_fp8
• weight_map::quantize_to_fp8_blockscaled

Type Aliases

• layers::glm5next_dsa::build::LoadFn
• layers::glm5next_mlp::build::ExpertFn
• layers::glm5next_mlp::build::LoadFn

Statics

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

Constants

• layer::VERIFY_WY_LAYER_STRIDE_BYTES
• layer::VERIFY_WY_TABLES_PER_LAYER
• layer::VERIFY_WY_TABLE_SEQS
• layer::VERIFY_WY_TABLE_STRIDE_BYTES
• layers::glm5next_dsa::DSA_MODULE
• layers::glm5next_dsa::KERNEL_KV_LORA_DIM
• layers::glm5next_dsa::KERNEL_MAX_KPOOL
• layers::glm5next_dsa::LAYERNORM_MODULE
• layers::glm5next_dsa::MASKED_ATTN_MAX_KEYS
• layers::glm5next_dsa::attend::DSA_DECODE_MODULE
• layers::glm5next_dsa::select::TOPK_SMEM_CEILING
• layers::glm5next_dsa_ref::INVALID
• layers::glm5next_kda::SMEM_CEILING
• layers::glm5next_kda::binding::DSA_MARKERS
• layers::glm5next_kda::binding::KDA_TENSORS
• layers::glm5next_layer::profile::DSA_ATTEND
• layers::glm5next_layer::profile::DSA_INDEXER
• layers::glm5next_layer::profile::DSA_PROJ
• layers::glm5next_layer::profile::DSA_SELECT
• layers::glm5next_layer::profile::KDA
• layers::glm5next_layer::profile::MHC
• layers::glm5next_layer::profile::MHC_POST
• layers::glm5next_layer::profile::MLP_DENSE
• layers::glm5next_layer::profile::MOE_COMBINE
• layers::glm5next_layer::profile::MOE_EXPERTS
• layers::glm5next_layer::profile::MOE_HOSTSYNC
• layers::glm5next_layer::profile::MOE_ROUTER
• layers::glm5next_layer::profile::MOE_SHARED
• layers::glm5next_layer::profile::NORM
• layers::glm5next_layer::profile::REDUCE_ATTN
• layers::glm5next_layer::profile::REDUCE_ATTN_BAR
• layers::glm5next_layer::profile::REDUCE_ATTN_ENQ
• layers::glm5next_layer::profile::REDUCE_MLP
• layers::glm5next_layer::profile::REDUCE_MLP_BAR
• layers::glm5next_layer::profile::REDUCE_MLP_ENQ
• layers::glm5next_mlp::FFN_MODULE
• layers::glm5next_mlp::GEMM_MODULE
• layers::glm5next_mlp::KERNEL_MAX_TOP_K
• layers::glm5next_mlp::W4A16_GEMV_MODULE
• layers::glm5next_mlp::W4A16_MODULE
• layers::glm5next_mlp::forward::MOE_ROW_BATCH_MAX_ROWS
• layers::glm5next_mlp::forward::MOE_ROW_UNION_MAX_IDS
• layers::ops::DENSE_GEMV_BATCHM_DECODE_MAX_M
• layers::ops::DENSE_GEMV_BATCHM_MAX_M
• layers::ops::GLM5NEXT_MHC_MODULE
• layers::ops::MAX_DYNAMIC_SMEM
• layers::ops::MHC_MIX_MAX_TOKENS
• layers::ops::MOE_TOPK_SIGMOID_MAX_EXPERTS
• layers::ops::MOE_TOPK_SIGMOID_MAX_TOP_K
• layers::ops::NVFP4_BLOCK_BYTES
• layers::ops::NVFP4_MMQ_SMEM
• layers::ops::Q2_0_BLOCK_BYTES
• layers::ops::Q2_BATCHM_MAX_M
• layers::ops::Q4K_BLOCK_BYTES
• layers::ops::Q4K_MMQ_SMEM
• 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::moe_lora_grouped::MLG_M_TILE
• layers::qwen3_ssm::gdn_flags::MAX_F16_TWIN_DFLASH_GAMMA
• layers::qwen3_ssm::gdn_flags::MAX_F16_TWIN_K
• layers::vision_encoder::IMAGE_PAD_TOKEN_ID
• layers::vision_encoder::VIDEO_PAD_TOKEN_ID
• layers::w4a16_gemv_tiers::W4A16_BATCHM_WIDTHS
• lora::ROWDIFF_THRESH
• speculative::EP_CMD_MTP_PROPOSE
• speculative::tree_shape::MAX_NODES
• speculative::tree_shape::MAX_RANK
• speculative::verify_key::CANONICAL_KEY_MIN_WIDTH
• video_preprocess::DEFAULT_FPS
• video_preprocess::DEFAULT_MAX_FRAMES
• video_preprocess::DEFAULT_MIN_FRAMES