List of all items

Structs

• arch::ArchMismatch
• arch::SmArch
• capabilities::ModelCapabilities
• compute::NvidiaTarget
• config::GgufConfigInputs
• config::ModelConfig
• config::PeftAdapterConfig
• config::QuantizationConfig
• config::VisionConfig
• cuda_host::CudaHost
• device::AvarokDevice
• dtype::QuantConfig
• fault::FaultLatch
• kernel::KernelModule
• kimi_k3::attnres::AttnResHub
• kimi_k3::cache::HybridCache
• kimi_k3::cache::MlaKv
• kimi_k3::cpu_forward::AttnResStream
• kimi_k3::cpu_forward::K3LayerCtx
• kimi_k3::cpu_weights::Ablation
• kimi_k3::cpu_weights::DenseMlp
• kimi_k3::cpu_weights::K3CpuLayer
• kimi_k3::cpu_weights::K3CpuModel
• kimi_k3::cpu_weights::KdaWeights
• kimi_k3::cpu_weights::MlaWeights
• kimi_k3::cpu_weights::MoeWeights
• kimi_k3::expert_backend::MmapExpertStore
• kimi_k3::expert_backend::PrefetchPlanner
• kimi_k3::kda::KdaConfig
• kimi_k3::kda::KdaState
• kimi_k3::latent_moe::LatentMoeConfig
• kimi_k3::layer::K3Graph
• kimi_k3::layer::K3LayerSpec
• kimi_k3::mla::MlaConfig
• kimi_k3::tp::TpBytePlan
• registry::AvarokRegistry
• registry::RawCudaFunc
• safetensors::TensorSpan
• scope::Teardown
• stream::AvarokStream
• target::KernelTarget
• tensor::TensorRef

Enums

• arch::SmSuffix
• capabilities::AttentionType
• capabilities::SsmArchitecture
• compute::Vendor
• config::Glm5NextRouterMode
• config::LayerType
• dtype::DType
• error::AvarokError
• fault::Fatality
• kimi_k3::cache::LayerCache
• kimi_k3::cpu_weights::MixerW
• kimi_k3::cpu_weights::MlpW
• kimi_k3::expert_backend::ExpertBackendKind
• kimi_k3::layer::MixerKind
• kimi_k3::layer::MlpKind
• kimi_k3::tp::TpAxis

Traits

• compute::ComputeTarget
• config::GgufMeta
• scope::ModelResource

Functions

• arch::parse_sm_arch
• arch::ptx_arch_runs_on_device
• arch::target_hint
• compute::target_for_vendor
• config::allow_partial_targets
• config::config_from_gguf
• config::glm5_next_mtp_layer_index
• config::glm_vision_enabled
• config::glm_vision_enabled_from
• config::parse_config
• config::parse_mistral_params
• config::parse_peft_adapter_config
• config::parse_quantization_config
• cuda_host::host
• cuda_host::release
• env_compat::mirror_legacy_env
• fault::classify
• fault::exit_code
• fault::global
• kernel::launch_config
• kernel::launch_vector_add
• kimi_k3::assemble_layer
• kimi_k3::attnres::attnres_blend
• kimi_k3::attnres::attnres_mix
• kimi_k3::attnres::attnres_softmax_mix
• kimi_k3::attnres::rms_norm
• kimi_k3::binding_memory::engine_bf16_weight
• kimi_k3::binding_memory::extra_gpu_bytes
• kimi_k3::cpu_forward::forward_one_layer
• kimi_k3::cpu_forward::forward_one_layer_with_cores
• kimi_k3::cpu_forward::forward_one_layer_with_kda_decode
• kimi_k3::cpu_forward::forward_one_layer_with_mla_decode
• kimi_k3::cpu_forward::forward_token
• kimi_k3::cpu_forward::logits
• kimi_k3::expert_backend::kind_from_env
• kimi_k3::expert_backend::pack_path
• kimi_k3::expert_backend::read_pack_file
• kimi_k3::expert_backend::write_dummy_pack
• kimi_k3::greedy::greedy_decode
• kimi_k3::kda::bounded_gate
• kimi_k3::kda::conv_update
• kimi_k3::kda::cuda_kda_enabled
• kimi_k3::kda::full_rank_output_gate
• kimi_k3::kda::kda_decode_token
• kimi_k3::kda::kda_from
• kimi_k3::kda::kda_recurrent_step
• kimi_k3::latent_moe::expert_situ
• kimi_k3::latent_moe::latent_moe_forward
• kimi_k3::latent_moe::mix_routed_experts
• kimi_k3::latent_moe::moe_from
• kimi_k3::latent_moe::sigmoid_topk
• kimi_k3::mla::apply_output_gate
• kimi_k3::mla::cuda_mla_enabled
• kimi_k3::mla::gated_mla_attend
• kimi_k3::mla::maybe_rope
• kimi_k3::mla::mla_decode_token
• kimi_k3::mla::mla_from
• kimi_k3::mla::sdpa
• kimi_k3::mla::sdpa_one
• kimi_k3::mxfp4::allow_mxfp4
• kimi_k3::mxfp4::is_expert_packed
• kimi_k3::mxfp4::is_expert_scale
• kimi_k3::mxfp4::scale_name
• kimi_k3::mxfp4::unpack_expert
• kimi_k3::mxfp4::unpack_expert_bf16
• kimi_k3::mxfp4::unpacked_name
• kimi_k3::ops::argmax
• kimi_k3::ops::embed_token
• kimi_k3::ops::fill
• kimi_k3::ops::ident
• kimi_k3::ops::matvec
• kimi_k3::ops::matvec_column_tp
• kimi_k3::ops::ones
• kimi_k3::situ::sigmoid
• kimi_k3::situ::silu
• kimi_k3::situ::situ_gate
• kimi_k3::situ::situ_glu
• kimi_k3::situ::situ_glu_vec
• kimi_k3::situ::situ_up
• kimi_k3::situ::softcap
• kimi_k3::situ::swiglu_mutant
• kimi_k3::text_key
• kimi_k3::tp::plan_tensor_bytes
• kimi_k3::tp::tensor_plan
• kimi_k3::weights::layer_keys
• kimi_k3::weights::required_names
• kimi_k3::weights::text_key
• kimi_k3::weights_geometry::expected_replicated_shape
• kimi_k3::weights_geometry::validate_replicated_shape
• mxfp4_e8m0::dequant_nvfp4_e8m0_to_bf16
• mxfp4_e8m0::dequant_nvfp4_e8m0_to_f32
• mxfp4_e8m0::fp8_e8m0_to_f32
• numeric::bf16_bytes_to_f32
• numeric::f32_to_bf16
• numeric::fp8_e4m3_to_f32
• registry::cuda_error_text
• registry::is_teardown_noop
• safetensors::tensor_span

Type Aliases

• error::Result
• kimi_k3::cpu_forward::DenseMlpCore
• kimi_k3::cpu_forward::HiddenReduce

Statics

• numeric::FP8_E4M3_LUT

Constants

• config::PEFT_SUPPORTED_TARGET_MODULES
• device::sm121::COMPUTE_MAJOR
• device::sm121::COMPUTE_MINOR
• device::sm121::MAX_REGS_PER_THREAD
• device::sm121::MAX_THREADS_PER_BLOCK
• device::sm121::MEMORY_BW_GBS
• device::sm121::NUM_SMS
• device::sm121::SMEM_PER_SM
• device::sm121::WARP_SIZE
• fault::EXIT_GPU_FAULT
• kimi_k3::expert_backend::ENV
• kimi_k3::kda::KDA_L2_EPS
• kimi_k3::mxfp4::PACKED_CLASSES
• mxfp4_e8m0::E2M1
• mxfp4_e8m0::GROUP_SIZE
• registry::CUDA_ERROR_DEINITIALIZED