MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 1 callersFunctionaddLanguageSwitcher
* Add a compact language switcher to the docs header. * * @returns {void}
docs/source/_static/custom.js:92
↓ 1 callersFunctionaddRunLlmWidget
* Add the RunLLM assistant widget script once per page. * * @returns {void}
docs/source/_static/custom.js:6
↓ 1 callersMethodadd_adapter_state
Register a new adapter's eviction state at runtime.
lmcache/v1/distributed/storage_controllers/eviction_controller.py:240
↓ 1 callersMethodadd_arguments
Add command-specific arguments to *parser*. Args: parser: The ``ArgumentParser`` for this subcommand.
lmcache/cli/commands/base.py:58
↓ 1 callersMethodadd_arguments
(self, parser: argparse.ArgumentParser)
lmcache/cli/commands/bench/engine_bench/__init__.py:24
↓ 1 callersMethodadd_arguments
Add benchmark-specific arguments. Only the torch-free ``config`` module is imported here, so registering this tool does not require t
lmcache/cli/commands/tool/transfer_channel_benchmark.py:30
↓ 1 callersMethodadd_arguments
(self, parser: argparse.ArgumentParser)
lmcache/cli/commands/tool/cache_simulator/sweep_command.py:25
↓ 1 callersMethodadd_arguments
(self, parser: argparse.ArgumentParser)
lmcache/cli/commands/tool/cache_simulator/simulate_command.py:26
↓ 1 callersMethodadd_arguments
(self, parser: argparse.ArgumentParser)
lmcache/cli/commands/tool/cache_simulator/gen_dataset_command.py:26
↓ 1 callersFunctionadd_cache_misses
Add cache misses to the prompts and return a boolean mask aligned with prompts.
benchmarks/long_doc_qa/long_doc_qa.py:371
↓ 1 callersFunctionadd_engine_arguments
Add ``lmcache bench engine`` arguments to *parser*. Args: parser: The ``ArgumentParser`` for the engine bench subcommand.
lmcache/cli/commands/bench/engine_bench/command.py:57
↓ 1 callersFunctionadd_header
(file_path)
tools/check_spdx_header.py:25
↓ 1 callersMethodadd_key_to_dict
( self, key: CacheEngineKey, obj: MemoryObjMetadata, index: int )
lmcache/v1/storage_backend/nixl_storage_backend.py:1076
↓ 1 callersMethodadd_l1_storage
L1 cache capacity, usage, eviction, and object count.
lmcache/cli/commands/describe.py:127
↓ 1 callersMethodadd_l2_adapters
L2 adapter sections.
lmcache/cli/commands/describe.py:255
↓ 1 callersFunctionadd_l2_arguments
Add ``lmcache bench l2`` arguments to *parser*. Args: parser: The ``ArgumentParser`` for the L2 bench subcommand.
lmcache/cli/commands/bench/l2_adapter_bench/command.py:29
↓ 1 callersMethodadd_models
Per-model KV cache layout sections. Each model gets one section with context-wide fields, followed by one ``kernel_groups`` list entr
lmcache/cli/commands/describe.py:173
↓ 1 callersMethodadd_overview
Top-level engine overview.
lmcache/cli/commands/describe.py:120
↓ 1 callersFunctionadd_replay_arguments
Add ``lmcache trace replay`` arguments to *parser*. Args: parser: The ``ArgumentParser`` for the replay subcommand.
lmcache/cli/commands/trace/replay_command.py:43
↓ 1 callersFunctionadd_server_arguments
Add ``lmcache bench server`` arguments to *parser*. Requires the full LMCache install (torch, zmq, etc.). Callers should check ``_IMPORT_ERRO
lmcache/cli/commands/bench/server_bench/command.py:84
↓ 1 callersMethodadd_sync_handler
( self, request_type: RequestType, payload_clss: list[Any], handler )
lmcache/v1/multiprocess/mq.py:739
↓ 1 callersMethodadd_worker
(worker_id)
tests/v1/cache_controller/test_registry_tree.py:175
↓ 1 callersFunctionaggregate_results
Aggregate results from multiple processes
lmcache/tools/controller_benchmark/__main__.py:43
↓ 1 callersFunctionall_derangements
Permutations p of range(n) with p[i] != i for all i (only for small n).
benchmarks/multi_doc_qa/shuffle_doc_qa.py:51
↓ 1 callersMethodall_loads_done
(self)
lmcache/v1/distributed/storage_controllers/prefetch_controller.py:162
↓ 1 callersMethodall_lookups_done
(self)
lmcache/v1/distributed/storage_controllers/prefetch_controller.py:159
↓ 1 callersMethodalloc
Allocate pinned memory and return the raw pointer.
lmcache/v1/memory_management.py:427
↓ 1 callersFunctionalloc_pinned_ptr
Non-CUDA equivalent of allocating pinned memory and returning pointer to it. On XPU, uses pin_memory=True (SYCL USM host allocation) for fast
lmcache/python_ops_fallback.py:420
↓ 1 callersMethodallocate
( self, shapes: Union[torch.Size, list[torch.Size]], dtypes: Union[torch.dtype, list[t
lmcache/v1/memory_management.py:2432
↓ 1 callersMethodallocate
Allocate memory object with memory allocator. Use LRU evictor if eviction is enabled.
lmcache/v1/storage_backend/storage_manager.py:327
↓ 1 callersMethodallocate
(self, size)
lmcache/v1/storage_backend/connector/eic_connector.py:81
↓ 1 callersFunctionallocate_and_copy_objects
Allocate the memory objects and copy the data from src_memory_objs to the newly allocated memory objects Args: allocator_backend
lmcache/v1/storage_backend/storage_manager.py:64
↓ 1 callersFunctionanalyze_chunk_hashes
Analyze chunk hash records and compute statistics. Args: records: List of chunk hash records top_n: Number of top reused chunks t
examples/chunk_statistics/analyze_chunk_hashes.py:51
↓ 1 callersFunctionanalyze_hit_rates_across_pool_sizes
( token_sequences: List[List[int]], pool_sizes_gb: List[Union[int, float, str]], tokens_per_gb: in
examples/agents/prefix_analysis.py:234
↓ 1 callersFunctionannotate
Dummy decorator when nvtx is not available.
lmcache/utils.py:23
↓ 1 callersFunctionappendInitContainerIfAbsent
appendInitContainerIfAbsent appends c to initContainers unless one of the same name is already present. Returns the slice.
operator/internal/webhook/pod_injector.go:364
↓ 1 callersFunctionappendVolumeIfAbsent
appendVolumeIfAbsent appends v to volumes unless a volume of the same name is already present (idempotency within a single Handle call). Returns the s
operator/internal/webhook/pod_injector.go:353
↓ 1 callersFunctionappendVolumeMountIfAbsent
appendVolumeMountIfAbsent appends m to mounts unless one of the same name is already present. Returns the slice.
operator/internal/webhook/pod_injector.go:378
↓ 1 callersMethodapply
Return the edited view for a layer this rule matched. Args: spec: The layer's vLLM KV cache spec (from its group). kv
lmcache/integration/vllm/kv_cache_group_edits.py:187
↓ 1 callersFunctionapply_kv_cache_group_edits
Apply all KV cache group metadata edits for LMCache registration. Each layer is checked against the ``_EDITS`` rules (first match wins) and r
lmcache/integration/vllm/kv_cache_group_edits.py:359
↓ 1 callersMethodapply_loaded_state
Validate and apply a recovered metadata checkpoint payload. Args: data: Decoded checkpoint dictionary. Returns:
lmcache/v1/storage_backend/raw_block/core.py:894
↓ 1 callersFunctionargValueLast
argValueLast returns the value following the LAST occurrence of flag. argparse-style flag handling lets a later --flag override an earlier one, so the
operator/test/e2e/smoke_helpers_test.go:236
↓ 1 callersFunctionassertGoldenKvTransferConfig
assertGoldenKvTransferConfig compares the on-cluster ConfigMap data to the embedded golden snapshot. The placeholder __NAMESPACE__ in the golden file
operator/test/e2e/crd_smoke_test.go:223
↓ 1 callersFunctionassertHasArg
(t *testing.T, args []string, flag string)
operator/internal/resources/resources_test.go:1081
↓ 1 callersFunctionassert_is_vllm_mla_or_flash_attn_or_flash_infer
Ensure that we have an Engine KV Cache Format that is either vLLM's MLA, flash attention, or flash infer. Accepted formats: - ``
lmcache/v1/gpu_connector/utils.py:365
↓ 1 callersMethodasync_batched_read
Read a batch of data through the channel. :param buffers: A list of bytes or MemoryObj to store the read data. :param transf
lmcache/v1/transfer_channel/nixl_channel.py:550
↓ 1 callersMethodasync_batched_submit_put_task
( self, keys: Sequence[CacheEngineKey], objs: List[MemoryObj], transfer_spec:
lmcache/v1/storage_backend/p2p_backend.py:662
↓ 1 callersMethodasync_lookup_and_prefetch
Perform asynchronous lookup and prefetching across all storage backends. :param str lookup_id: The unique id (e.g., request id) for
lmcache/v1/storage_backend/storage_manager.py:651
↓ 1 callersMethodasync_try_admit
Async (asyncio) admission for receiver. :param req_id: The request identifier to admit. :param total_chunks: Number of chunks to rese
lmcache/v1/storage_backend/pd_backend_async.py:138
↓ 1 callersMethodbatch_exists
(self, keys: list)
tests/conftest.py:330
↓ 1 callersMethodbatch_exists_sync
Check existence of multiple keys in a single batch operation (sync version).
lmcache/v1/storage_backend/resp_client.py:207
↓ 1 callersMethodbatch_get
(self, keys: list[str], bufs: list[memoryview])
lmcache/v1/storage_backend/resp_client.py:173
↓ 1 callersMethodbatch_get_sync
(self, keys: list[str], bufs: list[memoryview])
lmcache/v1/storage_backend/native_clients/connector_client_base.py:139
↓ 1 callersMethodbatch_set
(self, keys: list[str], bufs: list[memoryview])
lmcache/v1/storage_backend/resp_client.py:187
↓ 1 callersMethodbatch_set_sync
(self, keys: list[str], bufs: list[memoryview])
lmcache/v1/storage_backend/native_clients/connector_client_base.py:146
↓ 1 callersMethodbatched_allocate
( self, shapes: Union[torch.Size, list[torch.Size]], dtypes: Union[torch.dtype, list[t
lmcache/v1/memory_management.py:2858
↓ 1 callersMethodbatched_allocate
( self, shapes: Union[torch.Size, list[torch.Size]], dtypes: Union[torch.dtype, list[t
lmcache/v1/storage_backend/pd_backend.py:367
↓ 1 callersMethodbatched_async_contains
( self, lookup_id: str, keys: List[CacheEngineKey], pin: bool = False, )
lmcache/v1/storage_backend/connector/bigtable_connector.py:779
↓ 1 callersMethodbatched_contains
Check whether the key exists in the storage backend. :param List[CacheEngineKey] keys: The keys to check. :param Optional[L
lmcache/v1/storage_backend/storage_manager.py:967
↓ 1 callersMethodbatched_contains
Return the number of consecutive keys present in the index. Synchronous variant of :meth:`batched_async_contains`. Args:
lmcache/v1/storage_backend/plugins/dax_backend.py:550
↓ 1 callersMethodbatched_exists
Alias for batch_exists.
lmcache/v1/storage_backend/resp_client.py:214
↓ 1 callersMethodbatched_free
Batched free memory objs. Unlike `batched_allocate`, this function does not assume that the memory objs are equal-sized.
lmcache/v1/memory_management.py:2069
↓ 1 callersMethodbatched_free
( self, memory_objs: List[MemoryObj], allocator_type: Optional[str] = None, up
lmcache/v1/memory_management.py:2934
↓ 1 callersMethodbatched_from_gpu
This function is a generator that moves the KV cache from the paged GPU memory to the memory objects. The first iteration will prepar
lmcache/v1/gpu_connector/gpu_connectors.py:938
↓ 1 callersMethodbatched_from_gpu
This function is a generator that moves the KV cache from the paged GPU memory to the memory objects. The first iteration will prepar
lmcache/v1/gpu_connector/xpu_connectors.py:759
↓ 1 callersMethodbatched_get
( self, keys: List[CacheEngineKey] )
lmcache/v1/storage_backend/connector/mock_connector.py:70
↓ 1 callersMethodbatched_get
Get multiple keys in parallel. Metrics reported per individual get.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:724
↓ 1 callersMethodbatched_get
( self, keys: List[CacheEngineKey] )
lmcache/v1/storage_backend/connector/s3_connector.py:406
↓ 1 callersMethodbatched_get_blocking
( self, keys: List[CacheEngineKey], )
lmcache/v1/storage_backend/remote_backend.py:396
↓ 1 callersMethodbatched_get_msg
Get a batch of messages from the message queue.
lmcache/v1/cache_controller/worker.py:387
↓ 1 callersMethodbatched_get_non_blocking
Retrieve memory objects for consecutive keys asynchronously. Schedules one batched restore job on the persistent dispatch executor an
lmcache/v1/storage_backend/plugins/dax_backend.py:479
↓ 1 callersMethodbatched_put
Upload multiple full chunks in a single Hugging Face batch call.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:300
↓ 1 callersMethodbatched_put_callback
Callback function for batched put tasks.
lmcache/v1/storage_backend/remote_backend.py:274
↓ 1 callersMethodbatched_recv
( self, buffers: Union[list[bytes], list[MemoryObj]], transfer_spec: Optional[dict] =
lmcache/v1/transfer_channel/nixl_channel.py:393
↓ 1 callersMethodbatched_send
( self, objects: Union[list[bytes], list[MemoryObj]], transfer_spec: Optional[dict] =
lmcache/v1/transfer_channel/nixl_channel.py:386
↓ 1 callersMethodbatched_submit_put_task
Submit batched put tasks to store KV caches to remote storage. :param on_complete_callback: Optional callback invoked once per key
lmcache/v1/storage_backend/remote_backend.py:281
↓ 1 callersMethodbatched_submit_put_task
P2P backend does not support put operations.
lmcache/v1/storage_backend/p2p_backend.py:755
↓ 1 callersMethodbatched_to_gpu
( self, memory_objs: Union[ List[List[MemoryObj]], List[MemoryObj], List[int], Non
lmcache/v1/gpu_connector/musa_connectors.py:806
↓ 1 callersMethodbatched_to_gpu
This function is a generator that moves the KV cache from the memory objects to buffer GPU memory. In each iteration i, it (1) loads
lmcache/v1/gpu_connector/gpu_connectors.py:778
↓ 1 callersMethodbatched_to_gpu
This function is a generator that moves the KV cache from the memory objects to buffer GPU memory. In each iteration i, it (1) loads
lmcache/v1/gpu_connector/xpu_connectors.py:602
↓ 1 callersFunctionbench_load
Benchmark ``submit_load_task``.
lmcache/cli/commands/bench/l2_adapter_bench/runner.py:274
↓ 1 callersFunctionbench_lookup
Benchmark ``submit_lookup_and_lock_task``.
lmcache/cli/commands/bench/l2_adapter_bench/runner.py:204
↓ 1 callersFunctionbench_store
Benchmark ``submit_store_task``. For each round, ``in_flight`` independent submits are issued; the round duration is the wall-clock time from
lmcache/cli/commands/bench/l2_adapter_bench/runner.py:134
↓ 1 callersMethodblend
Perform blending for the given tokens.
lmcache/v1/compute/blend/blender.py:153
↓ 1 callersMethodblend_layer
Perform layerwiese retrieve + blending.
lmcache/v1/compute/blend/blender.py:125
↓ 1 callersMethodblock_size
Return the block size (tokens per block; raises for NBBS-fused).
lmcache/v1/gpu_connector/kv_format/specs/base.py:123
↓ 1 callersFunctionblocked_req
()
tests/v1/storage_backend/test_pd_backend_async.py:1068
↓ 1 callersFunctionboolPtr
boolPtr returns a pointer to b (local helper to avoid importing a ptr pkg).
operator/test/e2e/coordinator_registration_smoke_test.go:134
↓ 1 callersFunctionboom
(method: str, path: str, payload: dict)
tests/v1/mp_coordinator/test_blend_client.py:168
↓ 1 callersMethodbucket_info
Return bucket metadata for auth and health checks.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:135
↓ 1 callersFunctionbuildL2AdapterJSON
buildL2AdapterJSON serializes an L2BackendSpec into the --l2-adapter JSON string. For RESP adapters with authSecretRef, username/password are set to e
operator/internal/resources/compute.go:141
↓ 1 callersFunctionbuildRESPL2JSON
(resp *lmcachev1alpha1.RESPL2AdapterSpec)
operator/internal/resources/compute.go:151
↓ 1 callersFunctionbuildRawL2JSON
(raw *lmcachev1alpha1.RawL2AdapterSpec)
operator/internal/resources/compute.go:171
↓ 1 callersFunctionbuild_argument_parser
()
benchmarks/long_doc_qa/long_doc_qa_recommender.py:186
↓ 1 callersFunctionbuild_common_cpp
Build pure C++ extensions that do not depend on any backend. Args: profile: Resolved backend profile (or ``None``). Each spec in
setup_extensions/common_cpp.py:67
↓ 1 callersFunctionbuild_config
Build a :class:`BenchmarkConfig` from parsed arguments. Args: args: Parsed CLI arguments produced by ``add_benchmark_arguments``. Re
lmcache/tools/transfer_channel_benchmark/config.py:225
↓ 1 callersFunctionbuild_documents
Synthetic docs aligned with multi_doc_qa.py.
benchmarks/multi_doc_qa/shuffle_doc_qa.py:99
↓ 1 callersFunctionbuild_llm
(model: str)
tests/v1/compute/attention/test_triton_sparse_e2e.py:51
↓ 1 callersFunctionbuild_llm_with_lmcache
Build a vLLM LLM instance with LMCache integration. Creates a context manager that builds a vLLM LLM instance configured with LMCache for KV
examples/kv_cache_reuse/local_backends/rust_backend_offload.py:63
↓ 1 callersFunctionbuild_llm_with_lmcache
(lmcache_connector: str, model: str, vllm_version: str)
examples/kv_cache_reuse/local_backends/offload.py:47
← previousnext →3,801–3,900 of 13,371, ranked by callers