List of all items
Structs
- Layout
- Shape
- Storage
- decode_shape::KvAllocId
- decode_shape::ModelInstanceId
- decode_shape::ShapeKeyHasher
- device::Device
- factories::CpuFactory
- inference_context::DecodeSession
- inference_context::DecodeTokenData
- inference_context::InferenceContext
- inference_context::KvCache
- inference_context::KvLayer
- inference_context::KvLayerId
- inference_context::LatentKvCache
- inference_context::LatentSlot
- inference_context::PagedDecodeSession
- inference_context::PagedDecodeTokenData
- judge::Judge
- judge::ProfileEntry
- judge::ProfileJudgeOracle
- judge::ProfileReport
- judge::SizeClass
- judge::cache::DispatchOptions
- judge::cache::DispatchTable
- judge::cache::Pick
- judge::cache::ProfileEntry
- judge::cache::ProfileReport
- judge::cache::SizeClass
- kv_block_pool_device::DeviceEvicted
- kv_block_pool_device::DeviceKvPool
- kv_block_pool_device::PageTableHost
- lazy::LayerWeights
- lazy::Tensor
- lazy::TokenDataHost
- lazy_latent_cache::LatentCache
- nf4::BnbNestedState
- nf4::BnbQuantState
- nf4::Nf4Weight
- persistent_decode::DecodeDims
- persistent_decode::DecodeLayerInputs
- persistent_decode::DecodeTokenHost
- persistent_decode::MaskPlan
- persistent_decode::RopePlan
- planner::Planner
- reproducibility::StrictReproducibilityGuard
- scheduling::ScheduleOptions
Enums
- D
- DType
- DTypeParseError
- DeviceLocation
- Error
- HostBuffer
- device::DeviceLocation
- dtype::DType
- dtype::DTypeParseError
- inference_context::AuthorityState
- inference_context::KvSlot
- inference_context::PagedDecodePlan
- inference_context::PlanValidity
- inference_context::RebindRefusal
- judge::OpKind
- judge::OpSize
- judge::cache::Criterion
- judge::cache::OpKind
- kv_block_pool_device::BlockKind
- lazy::SamplingStrategy
- lazy::SessionDisposition
- lazy::WeightStorage
Traits
- Context
- device::NdArray
- dtype::FloatDType
- dtype::IntDType
- dtype::WithDType
- factories::BackendFactory
- factories::Realizer
- inference_context::KvRebindSource
- persistent_decode::DecodeBackbone
- persistent_decode::PersistentDecodeModel
Macros
Functions
- cuda_backend::device_if_available
- cuda_backend::new_device
- cuda_backend::new_device_with_stream
- factories::factory_for
- factories::registry
- inference_context::alloc_batched_kv
- judge::cache::cached
- judge::cache::cached_oracle
- judge::cache::invalidate
- judge::cache::populate_dispatch_table
- judge::default_report_path
- kv_block_pool_device::fit_geometry_from_residency
- lazy::apply_affine_rms_norm
- lazy::build_decode_causal_mask
- lazy::build_decode_causal_mask_windowed
- lazy::flash_window_bounds
- lazy::invalidate_decode_pair_if_stale
- lazy::load_tensor_as_f32
- lazy::load_transposed_matrix
- lazy::load_transposed_matrix_preserve_dtype
- lazy::offer_flash_decode_arm_for_region
- lazy::realize_many_f32
- lazy::refresh_decode_session
- lazy::sample_logits
- lazy::sample_multinomial
- lazy::spec_argmax
- lazy::spec_next_u01
- lazy::spec_sample_cat
- lazy::spec_softmax_temp
- metal_backend::device_if_available
- metal_backend::new_device
- nf4::dequantize_nested_absmax
- nf4::load_nf4_layer
- nf4::nf4_from_bytes
- nf4::parse_bnb_quant_state
- persistent_decode::build_and_realize_first_decode_token
- persistent_decode::build_decode_mask_variants
- persistent_decode::compute_decode_token_host
- persistent_decode::drop_decode_session
- persistent_decode::forward_with_kv_context
- persistent_decode::forward_with_kv_context_persistent
- persistent_decode::rebind_and_realize_prebuilt
- persistent_decode::upload_decode_token_data
- pipelined_bridge::device_seed_storage
- pipelined_bridge::optimize_calls
- pipelined_bridge::optimize_calls_thread_local
- pipelined_bridge::prebuild_optimized_env
- pipelined_bridge::prebuild_optimized_env_capturing_cache
- pipelined_bridge::realize_many_as
- pipelined_bridge::realize_many_as_with_initial
- pipelined_bridge::realize_many_as_with_initial_env
- pipelined_bridge::realize_one_as
- pipelined_bridge::realize_one_as_multi_device
- pipelined_bridge::realize_one_as_with_initial
- pipelined_bridge::realize_one_as_with_initial_env
- pipelined_bridge::realize_one_as_with_initial_reporting
- pipelined_bridge::realize_one_bytes
- pipelined_bridge::realize_one_bytes_with_initial
- pipelined_bridge::realize_one_prebuilt_cached_env
- pipelined_bridge::realize_one_prebuilt_env
- pipelined_bridge::realize_one_reference_as
- pipelined_bridge::realize_split_as_with_initial
- pipelined_bridge::upload_host_buffer_to_device
- reproducibility::is_strict_reproducibility
- reproducibility::set_global_strict_reproducibility
- reproducibility::strict_reproducibility
- scheduling::apply_placement_plan
- scheduling::auto_place_and_route
- scheduling::auto_place_and_route_with_transfer_cost
- scheduling::dp_plan
- scheduling::force_apply_placement_plan
- scheduling::prepare_dispatch_table
- scheduling::prepare_dp_inputs
- scheduling::recommend_for_node
- scheduling::recommend_placement
- test_utils::assert_allclose_f32
- test_utils::assert_close_vec1