List of all items

Structs

• act_layers::LayerFfnActs
• act_layers::SwigluClamps
• act_layers::XieluLayers
• attn_gate::AttnGate
• attn_gate::AttnGateSpec
• attn_temperature::AttnTemperature
• attn_temperature::DeclaredTemperature
• bert_encoder::BertEncoder
• bert_encoder::BertHparams
• bert_encoder::BertLayer
• bert_encoder::QkLayerNorm
• bert_gguf_loader::EncoderSpec
• capability::ArchProfile
• capability::SwaPattern
• capability::UnauditedTriage
• chat_template::ChatTemplate
• chat_template::RenderOptions
• config::KdaConfig
• config::KimiHybridAttention
• config::LayerRopeParams
• config::MlaConfig
• config::MlaRopeConfig
• config::ModelConfig
• config::RopeFreqs
• decoder::AttnWeights
• decoder::Decoder
• decoder::GptOssLayer
• decoder::GptOssWeights
• decoder::LayerWeights
• decoder::MoeWeights
• decoder::kv_window::KvWindowPolicy
• deepseek_v4_budget::Dsv4Args
• deepseek_v4_budget::Dsv4AutoCost
• deepseek_v4_budget::Dsv4BudgetTooSmall
• deepseek_v4_budget::Dsv4LayerSizes
• deepseek_v4_budget::Dsv4PoolSizes
• deepseek_v4_budget::Dsv4WindowCtx
• deepseek_v4_budget::Dsv4WindowPool
• deepseek_v4_budget::FreeListAllocator
• deepseek_v4_budget::FreeListExhausted
• deepseek_v4_budget::UnknownCompressRatio
• deepseek_v4_decoder::DeepseekV4AttnWeights
• deepseek_v4_decoder::DeepseekV4CsaWeights
• deepseek_v4_decoder::DeepseekV4DecodeState
• deepseek_v4_decoder::DeepseekV4DecoderConfig
• deepseek_v4_decoder::DeepseekV4DecoderLayerWeights
• deepseek_v4_decoder::DeepseekV4DecoderWeights
• deepseek_v4_decoder::DeepseekV4LayerState
• deepseek_v4_decoder::DeepseekV4MoeFfnWeights
• device_budget::DeviceBudget
• draft_model::DraftModelSpeculator
• dry::DryBreakers
• dry::DryParams
• dry::DryRequest
• dry::DryVocabMissing
• embedding_model::EmbeddingModel
• encoder::PairSequence
• engine::DeepseekV4Engine
• engine::Glm52Engine
• engine::KimiEngine
• engine::MlaDecodeState
• engine::MlaDenseFfn
• engine::MlaEngine
• engine::MlaLayerWeights
• engine::MlaMoeFfn
• engine::MlaMoeRuntime
• execution_plan::ExecutionPlan
• execution_plan::FusedOpCaps
• execution_plan::MemoryPlan
• execution_plan::PlanGeometry
• gdn::Gdn
• gdn::GdnHparams
• gdn::RecurrentMask
• gemma4_engine::Gemma4AttnWeights
• gemma4_engine::Gemma4DecodeState
• gemma4_engine::Gemma4Engine
• gemma4_engine::Gemma4Hparams
• gemma4_engine::Gemma4LayerWeights
• gemma4_engine::Gemma4Weights
• glm52_decoder::Glm52DecodeState
• glm52_decoder::Glm52DecoderConfig
• glm52_decoder::Glm52DecoderLayerWeights
• glm52_decoder::Glm52DecoderWeights
• glm52_decoder::Glm52DenseFfnWeights
• glm52_decoder::Glm52MoeFfnWeights
• glm52_gguf_loader::Glm52DenseFfnWeights
• glm52_gguf_loader::Glm52FileMeta
• glm52_gguf_loader::Glm52GgufHparams
• glm52_gguf_loader::Glm52MoeFfnWeights
• glm_dsa::Glm52AttnState
• glm_dsa::Glm52AttnWeights
• glm_dsa::Glm52MlaConfig
• glm_dsa::IndexerConfig
• glm_dsa::IndexerWeights
• grammar::candidates::Candidate
• grammar::candidates::DecodedPiece
• grammar::element::GrammarElement
• grammar::element::RulePos
• grammar::json_schema::GrammarBuilder
• grammar::lazy::LazyState
• grammar::lazy::LazyTriggers
• grammar::lazy::TriggerPattern
• grammar::machine::Grammar
• grammar::parser::ParsedGrammar
• grammar::utf8::PartialUtf8
• grammar_sampler::GrammarSampler
• hadamard_fold::HadamardMeta
• hrm::HrmStreams
• hybrid_engine::HybridEngine
• hybrid_engine::HybridUnavailable
• hyper_connections::HyperConnectionHeadWeights
• hyper_connections::HyperConnectionPreWeights
• kda::KdaAttnWeights
• kda::KdaState
• kimi_decoder::DenseMlpWeights
• kimi_decoder::KimiDecodeState
• kimi_decoder::KimiDecoderConfig
• kimi_decoder::KimiDecoderLayerWeights
• kimi_decoder::KimiDecoderWeights
• kimi_gguf_loader::KimiGgufHparams
• kimi_loader::BlockResidualWeights
• kimi_loader::KimiExpertSource
• kimi_loader::KimiRealHparams
• kimi_loader::KimiStoredExpertLayout
• kimi_tokenizer::KimiTokenizer
• kimi_validate::KimiCheckpointReport
• kv_budget::ContextFit
• kv_budget::KvBudget
• kv_budget::KvBudgetError
• kv_budget::KvResidency
• kv_budget::KvShape
• latent_moe::KimiExpertWeights
• latent_moe::KimiLatentMoeWeights
• latent_moe::KimiMoeConfig
• layer_shapes::LayerShape
• lightning::Lightning
• lightning::LightningDecay
• loader::GgufExpertSource
• loader::StoredExpertLayout
• loader::StoredMatrixSpec
• lora::LoraAdapter
• lora::LoraPair
• lora::LoraTensor
• lora_attach::LoraAttached
• lora_attach::LoraSpec
• mamba1::Mamba1
• mamba1::Mamba1Hparams
• mamba2::Mamba2
• mamba2::Mamba2Hparams
• minimax_engine::MinimaxEngine
• minimax_engine::MinimaxUnavailable
• mla::MlaAttnWeights
• mla_arch::MlaArch
• mla_gguf_loader::Deepseek2Hparams
• mla_yarn::MlaYarn
• mtp_blocks::TrunkLayers
• norm_sites::NormSites
• norm_sites::StoredNorm
• parallel_dense_ffn::ParallelDenseFfn
• parallel_residual::ParallelResidual
• penalty_window::PenaltyWindow
• plamo2_ssm::Plamo2Ssm
• plamo2_ssm::Plamo2SsmHparams
• prefix_cache::PrefixCache
• prefix_cache::PrefixCacheStats
• prefix_cache::PrefixMatch
• rank_head::RankHead
• recurrent_engine::RecurrentEngine
• recurrent_engine::RecurrentUnavailable
• rerank_pooler::IdentityMismatch
• rerank_pooler::SplicedPooler
• residency_report::ResidencyAssumptions
• residency_report::ResidencyLine
• residency_report::ResidencyReport
• sampler_order::SamplerOrder
• sampling::RecommendedSampling
• sampling::RequestedSampling
• sampling::Sampler
• sampling::SamplingParams
• sampling::reasoning_budget::ReasoningBudget
• sampling::reasoning_budget::ReasoningBudgetPlan
• scalar_multipliers::DeclaredMultipliers
• scalar_multipliers::MultiplierDims
• scalar_multipliers::MultiplierSupport
• scalar_multipliers::ResolvedMultipliers
• shortconv::ShortConv
• skip_stream::SkipStream
• speculative::DraftBlock
• speculative::DraftDist
• speculative::PromptLookupSpeculator
• speculative::SpeculativeDecodeResult
• speculative::SpeculativeOptions
• sub_norms::SubNorms
• swa_geometry::RotaryWidths
• swa_geometry::SwaGeometry
• t5_engine::T5Engine
• t5_engine::T5Unavailable
• tokenizer::ByteTokenizer
• tokenizer::GgufBpeTokenizer
• tokenizer::GgufPlamo2Tokenizer
• tokenizer::GgufSpmTokenizer
• tokenizer::GgufUnigramTokenizer
• tokenizer::GgufWordPieceTokenizer
• tokenizer::NormalizerOptions
• tokenizer::StopTokens
• vision::VisionConfig
• vision::VisionEncoderLayerWeights
• vision::VisionEncoderWeights
• vision::VisionMergerWeights
• vl_engine::VlEngine
• vl_engine::VlProjectorPair
• vl_engine::VlUnavailable

Enums

• alibi::MaxAlibiBias
• attn_gate::GateAct
• attn_gate::GatePresence
• attn_gate::GateWidth
• attn_temperature::AttnTemperatureError
• attn_temperature::FloorSource
• bert_encoder::BertFfn
• bert_encoder::BertTopology
• capability::ArchPath
• capability::ArchScope
• capability::DecoderFamily
• capability::MemoryKind
• capability::QkNormStyle
• capability::SwaWindowOverride
• capability::TriageClass
• chat_template::BuiltinTemplate
• chat_template::TemplateError
• clamp_kqv::ClampError
• config::AttentionKind
• config::BatchWindow
• config::FfnActivation
• config::LayerAttentionKind
• config::RopeLayout
• decoder::ExpertBacking
• decoder::MultiSeqKv
• deepseek_v4_budget::LayerCompressor
• device_budget::BudgetBackend
• draft_model::VocabMismatch
• embedding_model::EmbedError
• encoder::EncodeError
• engine::MlaLayerFfn
• engine_factory::EngineSelectError
• engine_factory::SelectedEngineKind
• engine_factory::ServedEngine
• gdn::BetaAlpha
• glm52_decoder::Glm52LayerFfn
• grammar::element::GreType
• grammar::error::GrammarError
• grammar::json_schema::JsonValue
• grammar::json_schema::SchemaError
• grammar::lazy::TriggerStep
• grammar_sampler::ConstraintError
• grammar_sampler::MaskOutcome
• kimi_decoder::KimiLayerAttention
• kimi_decoder::KimiLayerFfn
• kimi_decoder::KimiLayerState
• kimi_loader::KimiLoadError
• kimi_tokenizer::KimiTokenizerError
• kimi_validate::KimiValidateError
• kv_budget::Ceiling
• kv_budget::ContextCap
• kv_budget::KvElem
• kv_budget::KvLayout
• latent_moe::KimiExpertBacking
• layer_loops::HrmStream
• layer_loops::LayerLoops
• layer_loops::LoopNorm
• layer_shapes::AttnShape
• layer_shapes::LayerShapes
• layer_shapes::ZeroKvLayer
• loader::LoadError
• lora::LoraError
• mamba1::DtBcNorm
• mla::MlaKvB
• mla_arch::MlaOutputHead
• mla_arch::QProjRule
• mla_q_proj::MlaQProj
• mla_yarn::YarnLogMul
• mrope::MropeOnText
• norm::NormFunction
• norm::NormOp
• norm::NormParam
• parallel_dense_ffn::DensePresence
• parallel_residual::ParallelNorm
• parallel_residual::ParallelWhen
• pooling::PoolingError
• pooling::PoolingType
• proj_bias::Presence
• rerank_pooler::SpliceError
• rope_layers::NoRopePhase
• rope_layers::RopeLayers
• router_input::RouterInput
• sampler_order::ChainStep
• sampler_order::SamplerName
• sampler_order::SamplerOrderError
• sampling::reasoning_budget::BudgetState
• scalar_multipliers::AttentionScaleKey
• scalar_multipliers::LogitScaleUse
• scalar_multipliers::MultiplierDefaults
• scalar_multipliers::MultiplierError
• scalar_multipliers::ResidualScaleUse
• ssm_block::SsmBlock
• swa_layers::PatternKeyRead
• swa_layers::SwaLayers
• tensor_role::TensorRole
• tokenizer::SpecialTokens
• tokenizer::TokenizerLoadError

Traits

• dry::DryVocab
• encoder::TextEncoder
• engine::Engine
• engine::TextTokenizer
• grammar::parser::GrammarVocab
• speculative::Drafter

Functions

• act_layers::clamp_form
• act_layers::read_swiglu_clamps
• act_layers::read_xielu_layers
• act_layers::reads_swiglu_clamps
• act_layers::uses_xielu
• alibi::max_alibi_bias
• alibi::positions_by_alibi
• attn_gate::apply_interleaved_gate
• attn_gate::attn_gate_spec
• attn_gate::q_gate_interleaved
• attn_gate::split_interleaved_q_gate
• attn_temperature::resolve_attn_temperature
• attn_value_scale::resolve_attn_value_scale
• bert_gguf_loader::check_arch
• bert_gguf_loader::load_bert_encoder
• bert_gguf_loader::load_bert_encoder_from_path
• bert_gguf_loader::read_bert_hparams
• block_residual::apply_attn_res
• block_residual::apply_attn_res_prescored
• capability::architecture_catalog
• capability::attention_scale_override
• capability::canonical_architecture
• capability::coverage_report_markdown
• capability::default_swa_layout
• capability::embeddings_scaled_by_sqrt_n_embd
• capability::is_audited_generic
• capability::is_post_norm_only
• capability::resolve_architecture
• capability::resolve_profile
• capability::swa_disabled_by_arch
• capability::swa_rope_base_follows_model
• capability::swa_rope_scale_follows_model
• capability::swa_window_override
• capability::unaudited_refusal_detail
• capability::unaudited_triage
• capability::unsupported_feature_keys
• capability::unsupported_scaling_keys
• capability::uses_biased_layer_norm
• capability::uses_biased_rms_norm
• capability::uses_geglu
• capability::uses_gelu_ungated
• capability::uses_non_parametric_layer_norm
• capability::uses_non_parametric_rms_norm
• capability::uses_per_head_distinct_qk_norm
• capability::uses_per_head_scalar_qk_gain
• capability::uses_reglu
• capability::uses_relu_sqr
• capability::uses_weighted_layer_norm
• chunked_swa::chunked_window
• clamp_kqv::clamp_in_place
• clamp_kqv::resolve_clamp
• config::deepseek_v4_pro
• config::glm_5_2
• config::kimi_k3
• config::test_dense_fixture
• config::test_mixed_fixture
• config::test_moe_fixture
• ctk::is_accepted
• ctk::is_llama_cpp_type
• ctk::is_served
• ctk::parse_value
• ctk::unsupported_message
• deepseek_v4_budget::dsv4_auto_cost_model
• deepseek_v4_budget::dsv4_cache_per_page
• deepseek_v4_budget::dsv4_kv_unit_bytes
• deepseek_v4_budget::dsv4_pool_bytes
• deepseek_v4_budget::dsv4_pool_sizes
• deepseek_v4_budget::dsv4_reserved_window_pages
• deepseek_v4_budget::dsv4_solve_num_pages
• deepseek_v4_budget::dsv4_window_floor_pages
• deepseek_v4_budget::dsv4_window_unit_bytes
• deepseek_v4_budget::ring_size_for_ratio
• deepseek_v4_budget::window_ring_position
• deepseek_v4_decoder::deepseek_v4_forward_token
• devices::print_available_devices
• embedding_model::is_embedding_arch
• engine_factory::ensure_generic_decoder
• engine_factory::load_gemma4_engine_from_path
• engine_factory::load_glm52_engine_from_path
• engine_factory::load_mla_engine_from_path
• engine_factory::select_engine_kind
• gdn::head_map
• gdn::recurrent_layers
• gemma4_gguf_loader::load_gemma4_engine
• gemma4_gguf_loader::read_gemma4_hparams
• glm52_decoder::glm52_forward_token
• glm52_gguf_loader::glm52_indexer_config
• glm52_gguf_loader::glm52_mla_config
• glm52_gguf_loader::load_glm52_attn
• glm52_gguf_loader::load_glm52_dense_ffn
• glm52_gguf_loader::load_glm52_engine
• glm52_gguf_loader::load_glm52_moe_ffn
• glm52_gguf_loader::read_glm52_hparams
• glm_dsa::glm52_attn_forward_token
• grammar::candidates::accepts_token
• grammar::candidates::decode_for
• grammar::candidates::reject_candidates
• grammar::json_schema::json_schema_to_grammar
• grammar::json_schema::json_schema_to_grammar_value
• grammar::parser::parse
• grammar::parser::parse_with_vocab
• hadamard_fold::fold_for
• hadamard_fold::meta_for
• hf_pull::looks_like_hf_repo
• hf_pull::pull_hf_gguf
• hf_pull::resolve_model_path
• hrm::hrm_schedule
• hyper_connections::head
• hyper_connections::post
• hyper_connections::pre
• hyper_connections::sinkhorn
• hyper_connections::weighted_sum
• kda::kda_forward_token
• kimi_decoder::kimi_forward_token
• kimi_generate::kimi_generate
• kimi_gguf_loader::load_kimi_gguf_checkpoint
• kimi_gguf_loader::load_kimi_gguf_layer
• kimi_loader::load_block_residual
• kimi_loader::load_dense_mlp
• kimi_loader::load_f32_vec
• kimi_loader::load_kda_attn
• kimi_loader::load_kimi_checkpoint
• kimi_loader::load_kimi_checkpoint_with_expert_cache
• kimi_loader::load_kimi_expert
• kimi_loader::load_kimi_layer
• kimi_loader::load_latent_moe
• kimi_loader::load_mla_attn
• kimi_tokenizer::parse_special_tokens
• kimi_tokenizer::parse_tiktoken_vocab
• kimi_validate::validate_kimi_checkpoint_dir
• kv_head_dims::admits_split_kv_head_dims
• kv_head_dims::resolve_v_head_dim
• latent_moe::kimi_latent_moe_forward
• layer_loops::read_layer_loops
• layer_shapes::per_layer_shapes_read_by_llama_cpp
• layer_shapes::pure_recurrent_block
• layer_shapes::read_u64_per_layer
• layer_shapes::read_u64_trunk_layers
• loader::assert_every_tensor_consumed
• mamba2::parallel_with_attention
• mla::mla_forward_token
• mla_arch::is_lite
• mla_arch::mla_arch
• mla_gguf_loader::load_mla_engine
• mla_gguf_loader::read_deepseek2_hparams
• mla_yarn::kq_scale
• mla_yarn::resolve_mla_yarn
• mmproj::eprint_mmproj_if_present
• mmproj::find_mmproj_beside
• mmproj::warn_mmproj_if_present
• moe_interleave::dense_by_router_absence
• moe_interleave::interleave_step
• moe_interleave::interleave_step_refusal
• mrope::declares_mrope
• mrope::mrope_refusal
• mtp_blocks::is_mtp_block_tensor
• mtp_blocks::note_mtp_blocks_skipped
• mtp_blocks::reads_nextn
• mtp_blocks::trunk_layers
• norm::norm_function
• norm::norm_function_for_file
• norm::post_norm_eps
• norm::rms_norm_no_params
• norm_sites::weightless_embedding_norm
• normed_residual::adopt
• output_projection::grouped_output_projection
• parallel_dense_ffn::parallel_dense_ffn
• parallel_dense_ffn::parallel_dense_for_layer
• parallel_dense_ffn::shared_expert_sum_scale
• parallel_residual::layer_is_parallel
• parallel_residual::layer_parallel_norm
• parallel_residual::model_has_parallel_layer
• parallel_residual::parallel_residual
• plamo2_ssm::dt_dim
• pooling::l2_normalize
• pooling::pool
• position_embd::learned_positions
• position_embd::load_position_embd
• proj_bias::load_attn_out_bias
• proj_bias::load_dense_ffn_bias
• proj_bias::load_output_bias
• qk_layer_norm::per_head_layer_norm_refusal
• qk_layer_norm::uses_per_head_layer_norm_qk
• rank_head::load_rank_head
• rerank_pooler::classifier_matches
• rerank_pooler::splice_pooler
• rope_finetuned::unrotated
• rope_layers::reads_rope_pattern
• rope_layers::rope_layers
• rope_ntk_alpha::ntk_alpha_scaled_rope_base
• routed_weight_site::routed_slot
• routed_weight_site::weight_before_ffn
• router_input::router_input
• sampling::reasoning_budget::lossy_piece_is_complete
• sampling::sampling_distribution
• scalar_multipliers::multiplier_support
• scalar_multipliers::residual_add
• scalar_multipliers::resolve
• shortconv::is_shortconv_architecture
• shortconv::window_from_history
• skip_stream::has_skip_stream
• skip_stream::load_out_scale
• speculative::accept_or_resample
• speculative::speculative_decode
• speculative::speculative_decode_observed
• speculative::speculative_decode_with
• sub_norms::block_sub_norms
• sub_norms::load_sub_norms
• swa_geometry::full_layers_rotate_half
• swa_geometry::rotary_widths
• swa_geometry::swa_geometry_refusal
• swa_geometry::swa_layers_drop_rope_factors
• swa_geometry::swa_layers_unscaled_rope
• swa_geometry::two_widths_with_factors_refusal
• swa_geometry::window_required
• swa_layers::array_length
• swa_layers::pattern_key_read
• swa_layers::read_swa_layers
• tokenizer::eog_token_ids
• tokenizer::prepend_bos
• tokenizer::should_add_bos_token
• unread_tensors::note_unread_layer_tensors
• unread_tensors::unread_layer_tensors
• vision::embed_patches
• vision::encoder_forward
• vision::patch_merge
• vision::project_merged_patches
• weight_scales::load_projection_gain
• weight_scales::refuse_weight_scale_tensors
• weightless_qk_norm::weightless_qk_norm
• yarn_magnitude::yarn_attn_magnitude
• yarn_magnitude::yarn_log_mul_for

Type Aliases

• grammar::element::GrammarRule
• grammar::element::GrammarStack
• sampling::LogitMask

Constants

• act_layers::CLAMP_BEFORE_SILU
• act_layers::SWIGLU_CLAMP_READERS
• act_layers::XIELU_ARCHITECTURES
• act_layers::XIELU_KEYS
• alibi::ALIBI_ARCHITECTURES
• alibi::LLAMA_CPP_ALIBI_LITERAL
• attn_gate::ATTN_GATE_ARCHS
• attn_gate::GDN_Z_GATE_ARCHS
• attn_gate::Q_INTERLEAVED_GATE_ARCHS
• attn_temperature::ATTN_TEMPERATURE_READERS
• attn_temperature::LITERAL_ATTN_TEMPERATURE
• attn_value_scale::VALUE_SCALE_READERS
• bert_gguf_loader::BERT_ARCH
• bert_gguf_loader::ENCODER_ARCHS
• capability::AUDITED_GENERIC_GQA
• capability::BIASED_LAYER_NORM
• capability::BIASED_RMS_NORM
• capability::LOGIT_SOFTCAP_ARCHITECTURES
• capability::NON_PARAMETRIC_LAYER_NORM
• capability::NON_PARAMETRIC_RMS_NORM
• capability::PER_HEAD_DISTINCT_QK_NORM
• capability::PER_HEAD_SCALAR_QK_GAIN
• capability::POST_NORM_ONLY_ARCHITECTURES
• capability::SMALLTHINKER_PINNED_WINDOW
• capability::TRIAGE_PENDING
• capability::WEIGHTED_LAYER_NORM
• chunked_swa::CHUNKED_SWA_ARCHITECTURES
• clamp_kqv::CLAMPED_QKV_ARCHITECTURES
• clamp_kqv::CLAMP_KQV_REQUIRED
• config::MODEL_LEVEL_TENSORS_READ_BY_CONFIG
• ctk::FRINK_EXTRA_KV_CACHE_TYPES
• ctk::LLAMA_CPP_KV_CACHE_TYPES
• decoder::kv_window::KV_WINDOW_ENV
• deepseek_v4_budget::AUTO_KV_SLACK_BYTES
• deepseek_v4_budget::BF16_BYTES
• deepseek_v4_budget::CSA_RATIO
• deepseek_v4_budget::DEFAULT_WINDOW_PAGE
• deepseek_v4_budget::FP32_BYTES
• deepseek_v4_budget::HCA_RATIO
• deepseek_v4_budget::INT64_BYTES
• deepseek_v4_budget::NO_WINDOW_SLOT
• device_budget::BUDGET_ENV
• device_budget::CPU_RESERVE_FRACTION
• device_budget::DEVICE_RESERVE_FRACTION
• dry::DEFAULT_SEQUENCE_BREAKERS
• gdn::GROUPED_HEAD_ARCHITECTURES
• gdn::INTERVAL_RECURRENT_ARCHITECTURES
• gemma4_engine::GEMMA4_ARCHES
• grammar::parser::MAX_REPETITION_THRESHOLD
• hyper_connections::HC_MULT
• kv_budget::CTX_AUTO_GRANULARITY
• kv_head_dims::SPLIT_KV_HEAD_DIM_ARCHS
• layer_loops::LOOP_READERS
• layer_shapes::BLOCK_WITHOUT_FFN_KEEPS_ITS_OUTPUT
• layer_shapes::PER_LAYER_SHAPE_ARCHS
• layer_shapes::PURE_RECURRENT
• mamba2::PARALLEL_WITH_ATTENTION
• mla_arch::MLA_ENGINE_ARCHS
• moe_interleave::DENSE_LAYER_BY_ROUTER_ABSENCE
• moe_interleave::INTERLEAVE_STEP_HONOURED_BY_LOADER
• moe_interleave::INTERLEAVE_STEP_IS_REQUIRED
• mrope::MROPE_READERS
• mtp_blocks::NEXTN_READERS
• norm::NORM_BY_RMS_EPS_KEY
• norm::POST_NORM_EPS_LITERAL
• norm_sites::ATTN_NORM_2_FEEDS_ATTENTION
• norm_sites::EMBEDDING_NORM_ARCHITECTURES
• norm_sites::NO_OUTPUT_NORM
• norm_sites::ONE_NORM_PER_LAYER
• norm_sites::OUTPUT_NORM_UNDER_EMBEDDING_NAME
• norm_sites::POST_NORMS_UNDER_GROK_NAMES
• norm_sites::PRE_FFN_NORM_IS_ATTN_OUTPUT_NORM
• norm_sites::PRE_FFN_NORM_IS_POST_ATTENTION_NORM
• norm_sites::WEIGHTLESS_EMBEDDING_NORM
• normed_residual::NORMED_RESIDUAL_ARCHITECTURES
• parallel_dense_ffn::PARALLEL_DENSE_FFN_ARCHITECTURES
• parallel_dense_ffn::SHARED_EXPERT_SUM_SCALE
• parallel_residual::PARALLEL_RESIDUAL_GRAPHS
• position_embd::LEARNED_POSITION_CREATORS
• proj_bias::ATTN_OUT_BIAS_CREATORS
• proj_bias::FFN_BIAS_CREATORS
• proj_bias::OUTPUT_BIAS_CREATORS
• qk_layer_norm::PER_HEAD_LAYER_NORM_QK
• rerank_pooler::IDENTITY_TOLERANCE
• rerank_pooler::POOLER_SOURCE_KEY
• rerank_pooler::SPLICEABLE_HEAD_DTYPES
• rope_finetuned::ROPE_GATED_ON_FINETUNED
• rope_layers::PER_LAYER_ROPE_GATES
• rope_layers::ROPE_PATTERN_READERS
• rope_ntk_alpha::NTK_ALPHA_RESCALED_ROPE_BASE
• routed_weight_site::ROUTED_WEIGHT_BEFORE_FFN
• router_input::ROUTER_INPUT_TABLE
• shortconv::SHORTCONV_ARCHITECTURES
• skip_stream::SKIP_STREAM_ARCHS
• sub_norms::SUB_NORM_ARCHS
• swa_layers::ARRAY_AT_TRUNK_LENGTH
• swa_layers::PER_LAYER_ARRAY_READERS
• swa_layers::SCALAR_THEN_ARRAY_READERS
• unread_tensors::UNREAD_LAYER_TENSORS
• weight_scales::SERVED_SCALE_TENSORS
• yarn_magnitude::YARN_LOG_MUL_READERS