MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 1 callersMethodpop_completed_store_tasks
Drain and return completed store task results.
lmcache/v1/distributed/l2_adapters/raw_block_l2_adapter.py:444
↓ 1 callersMethodpop_completed_store_tasks
Pop all the completed store tasks with their results. Returns: dict[L2TaskId, L2StoreResult]: a dictionary mapping the t
lmcache/v1/distributed/l2_adapters/mock_l2_adapter.py:185
↓ 1 callersMethodpop_event
Pop and return the event with the given type and id.
lmcache/v1/event_manager.py:49
↓ 1 callersMethodpost_async
Non-blocking async post for WRITE operations.
lmcache/v1/storage_backend/nixl_storage_backend.py:664
↓ 1 callersMethodpost_init
Post initialization of the Registration Controller.
lmcache/v1/cache_controller/controllers/registration_controller.py:50
↓ 1 callersMethodpost_init
Post-initialization for standalone mode.
lmcache/v1/standalone/manager.py:64
↓ 1 callersMethodprebuilt_connection
(self)
lmcache/v1/storage_backend/connector/eic_connector.py:237
↓ 1 callersFunctionprecompute_all_kv
(config: PrecomputeConfig)
benchmarks/rag/precompute.py:60
↓ 1 callersMethodprepare_retrieve
Prepare retrieve. Returns chunks or shm views, or None on miss.
lmcache/v1/multiprocess/transfer_context/base.py:181
↓ 1 callersMethodprepare_retrieve
( self, key: IPCCacheServerKey, instance_id: int )
lmcache/v1/multiprocess/transfer_context/shm.py:183
↓ 1 callersMethodprepare_store
Prepare SHM buffers for a store operation. Returns: None: pickle mode — no pre-allocated buffers. Caller gathers all
lmcache/v1/multiprocess/transfer_context/base.py:152
↓ 1 callersMethodprepare_store
( self, key: IPCCacheServerKey, instance_id: int )
lmcache/v1/multiprocess/transfer_context/shm.py:145
↓ 1 callersFunctionprint_aggregated_results
Print aggregated benchmark results from all processes
lmcache/tools/controller_benchmark/__main__.py:100
↓ 1 callersFunctionprint_analysis
Print analysis results in a readable format. Args: analysis: Analysis results dictionary verbose: Whether to print detailed stati
examples/chunk_statistics/analyze_chunk_hashes.py:128
↓ 1 callersFunctionprint_lmcache_deployment_string
( model_name: str, tp: int, cpu_offload_GiB_per_gpu: float )
benchmarks/long_doc_qa/long_doc_qa_recommender.py:88
↓ 1 callersFunctionprint_long_doc_qa_workload_string
(model_name: str, tokens_in_offload_cache: int)
benchmarks/long_doc_qa/long_doc_qa_recommender.py:103
↓ 1 callersFunctionprint_memory_estimation
Print memory estimation results. Args: memory_est: Memory estimation dictionary
examples/chunk_statistics/analyze_chunk_hashes.py:302
↓ 1 callersMethodprint_results
Print benchmark results
lmcache/tools/controller_benchmark/benchmark.py:596
↓ 1 callersFunctionprint_statistics
(results: dict[str, Any])
lmcache/tools/cache_simulator/simulator.py:336
↓ 1 callersFunctionprint_summary_table
Print a summary comparison table.
benchmarks/microbenchmark/ttl_lock_benchmark.py:408
↓ 1 callersFunctionprint_time_series
Print time-series analysis results. Args: time_series: Time-series statistics dictionary
examples/chunk_statistics/analyze_chunk_hashes.py:272
↓ 1 callersFunctionprint_vllm_deployment_string
(model_name: str, tp: int)
benchmarks/long_doc_qa/long_doc_qa_recommender.py:78
↓ 1 callersFunctionprocess_chunk
Process a single chunk file from FSConnector. This function is called for each chunk file and performs: - Transf
lmcache/v1/internal_api_server/vllm/load_fs_chunks_api.py:249
↓ 1 callersMethodprocess_inbound
Process one inbound response from the server. Called by the shared ClientPollingLoop when the DEALER socket is readable. Only touche
lmcache/v1/multiprocess/mq.py:327
↓ 1 callersMethodprocess_outbound_task
(self)
lmcache/v1/multiprocess/mq.py:284
↓ 1 callersFunctionprocess_output
(filename)
benchmarks/multi_round_qa/multi-round-qa.py:689
↓ 1 callersMethodprocess_qkv
( self, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, residual: t
lmcache/v1/compute/blend/blender.py:60
↓ 1 callersFunctionprocess_single_prompt
Process a single prompt with the given client and model. Args: client: The OpenAI client for making API calls. model: The mo
benchmarks/long_doc_qa/long_doc_qa.py:214
↓ 1 callersFunctionprocess_single_prompt
Process a single prompt with the given client and model. Args: client: The OpenAI client for making API calls. model: The mo
benchmarks/multi_doc_qa/multi_doc_qa.py:101
↓ 1 callersMethodprocess_tokens
Process the tokens and return the corresponding cache engine keys. :param Union[torch.Tensor, List[int]] tokens: The tokens to process.
lmcache/v1/token_database.py:492
↓ 1 callersMethodprofile_broadcast
(self)
lmcache/observability.py:172
↓ 1 callersMethodprofile_from_gpu
(self)
lmcache/observability.py:217
↓ 1 callersMethodprofile_put
(self)
lmcache/observability.py:225
↓ 1 callersMethodprofile_to_gpu
(self)
lmcache/observability.py:180
↓ 1 callersFunctionprompt_bool
Prompt for a yes/no confirmation. Shows ``[Y/n]`` or ``[y/N]`` depending on the default. Enter alone accepts the default. Args:
lmcache/cli/commands/bench/engine_bench/interactive/terminal.py:112
↓ 1 callersFunctionproxy_request
Proxy requests to the specified target host and port or socket path. For security, non-socket targets must match a host/port already register
lmcache/lmcache_frontend/app.py:571
↓ 1 callersMethodpush
(self, index: int)
lmcache/v1/storage_backend/nixl_storage_backend.py:280
↓ 1 callersMethodput
( self, client_meta: ClientMetaMessage, kv_chunk_bytes: bytearray, )
lmcache/v1/server/storage_backend/local_backend.py:47
↓ 1 callersMethodput
Store a memory object in the local data dictionary.
lmcache/v1/storage_backend/pd_backend_async.py:1457
↓ 1 callersMethodput
Store data to ai-toolkit (queued with priority).
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:743
↓ 1 callersMethodput_callback
(self, future: Future, key: CacheEngineKey)
lmcache/v1/storage_backend/remote_backend.py:212
↓ 1 callersMethodput_one
Store one memory object in a DAX slot. This method exists for wrappers that need the existing non-MP single-object error semantics. S
lmcache/v1/storage_backend/dax/core.py:247
↓ 1 callersMethodput_reserved_from_ptr
Store one key by copying from an already reserved source pointer. Args: key: Destination key to commit. src_ptr: Sour
lmcache/v1/storage_backend/dax/core.py:403
↓ 1 callersMethodquery_deserialize_result
Query deserialization task completion. Returns True on success, False on failure, None if still pending. Non-idempotent: only returns
lmcache/v1/distributed/serde/base.py:191
↓ 1 callersMethodquery_load_result
(self, task_id: L2TaskId)
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:223
↓ 1 callersMethodquery_load_result
(self, task_id: L2TaskId)
lmcache/v1/distributed/l2_adapters/p2p_l2_adapter.py:337
↓ 1 callersMethodquery_load_result
(self, task_id: L2TaskId)
lmcache/v1/distributed/l2_adapters/nixl_store_dynamic_l2_adapter.py:462
↓ 1 callersMethodquery_load_result
(self, task_id: L2TaskId)
lmcache/v1/distributed/l2_adapters/serde_wrapper.py:292
↓ 1 callersMethodquery_lookup_and_lock_result
(self, task_id: L2TaskId)
lmcache/v1/distributed/l2_adapters/p2p_l2_adapter.py:251
↓ 1 callersMethodquery_lookup_and_lock_result
Return and remove a completed lookup result. Args: task_id: Adapter-local lookup task id. Returns: A bitmap
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:462
↓ 1 callersMethodquery_serialize_result
Query serialization task completion. Returns True on success, False on failure, None if still pending. Non-idempotent: only returns a
lmcache/v1/distributed/serde/base.py:163
↓ 1 callersFunctionrand_suffix
(n: int)
lmcache/v1/storage_backend/gds_backend.py:143
↓ 1 callersFunctionrandom_derangement
Uniform random derangement via shuffle-and-reject (small n only).
benchmarks/multi_doc_qa/shuffle_doc_qa.py:60
↓ 1 callersFunctionrandom_string
(N)
tests/v1/utils.py:200
↓ 1 callersMethodread_async
Enqueue a ``cuFileReadAsync`` on the stream. ``buf_base`` is the registered base pointer (e.g. ``buf.data_ptr()``). ``buf_offset`` is
lmcache/v1/gpu_connector/_cufile_async.py:306
↓ 1 callersFunctionread_context
()
examples/frontend/frontend.py:24
↓ 1 callersMethodreap_stale_instances
Reap non-GPU registrations that have gone silent. A ping-proven instance is judged against ``reap_timeout_s``; one that has never pin
lmcache/v1/multiprocess/modules/engine_driven_transfer.py:192
↓ 1 callersMethodreceive_all
(self, meta: ServerMetaMessage)
lmcache/v1/storage_backend/connector/lm_connector.py:57
↓ 1 callersMethodreconcileConnectionConfigMap
reconcileConnectionConfigMap creates or updates the <engine>-connection ConfigMap.
operator/internal/controller/cacheblend_reconcile_helpers.go:167
↓ 1 callersMethodreconcileCoordinatorService
reconcileCoordinatorService creates or updates the ClusterIP Service.
operator/internal/controller/coordinator_reconcile_helpers.go:98
↓ 1 callersMethodreconcileCoordinatorServiceMonitor
reconcileCoordinatorServiceMonitor converges the headless metrics Service and the ServiceMonitor when prometheus is enabled, and deletes them otherwis
operator/internal/controller/coordinator_reconcile_helpers.go:127
↓ 1 callersMethodreconcileDaemonSet
reconcileDaemonSet creates or updates the blend_v3 engine DaemonSet.
operator/internal/controller/cacheblend_reconcile_helpers.go:70
↓ 1 callersMethodreconcileDeployment
reconcileDeployment creates or updates the coordinator Deployment.
operator/internal/controller/coordinator_reconcile_helpers.go:67
↓ 1 callersMethodreconcileLookupService
reconcileLookupService creates or updates the node-local lookup Service.
operator/internal/controller/cacheblend_reconcile_helpers.go:112
↓ 1 callersMethodreconcileMetricsService
reconcileMetricsService creates or updates the headless metrics Service.
operator/internal/controller/cacheblend_reconcile_helpers.go:140
↓ 1 callersMethodreconcileRESPAuthSecret
reconcileRESPAuthSecret ensures a local copy of the RESP auth secret exists in the engine's namespace. It mirrors the LMCacheEngine path: a no-op unle
operator/internal/controller/cacheblend_reconcile_helpers.go:357
↓ 1 callersMethodreconcileServiceMonitor
reconcileServiceMonitor creates, updates, or deletes the ServiceMonitor.
operator/internal/controller/cacheblend_reconcile_helpers.go:194
↓ 1 callersMethodreconfigure
Apply one generic runtime reconfiguration operation. Args: operation: One of ``status``, ``add``, ``remove``, or ``resize``.
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:707
↓ 1 callersMethodreconfigure_l2_adapter
( self, adapter_index: int, operation: str, payload: dict[str, object], )
tests/v1/multiprocess/http_apis/test_reconfigure_api.py:57
↓ 1 callersMethodreconfigure_l2_adapter
( self, adapter_index: int, operation: str, payload: dict[str, object], )
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:42
↓ 1 callersMethodrecord_async
Record token IDs asynchronously. Args: token_ids: Token IDs to record (tensor or list) lookup_id: Unique identifier f
lmcache/v1/lookup_client/record_strategies/base.py:231
↓ 1 callersMethodrecord_failed_blocks
Record block IDs associated with failed load attempts. Args: request_id: request id from vLLM. expected_mask: Boolean
lmcache/integration/vllm/vllm_v1_adapter.py:894
↓ 1 callersMethodrecreate_backend
Close a backend and recreate it from current config. This is an atomic close-then-create operation that combines :meth:`clos
lmcache/v1/storage_backend/storage_manager.py:1306
↓ 1 callersMethodrecreate_lookup_client
Recreate the lookup client (close + create). Returns: dict: Result with old and new client type info.
lmcache/v1/manager.py:386
↓ 1 callersMethodrecreate_lookup_server
Recreate the lookup server (close + create). Returns: dict: Result with old and new server type info.
lmcache/v1/manager.py:454
↓ 1 callersMethodrecv_request
Receive a request from a client. The transport deserializes each data frame into a Python object before returning. Returns:
lmcache/v1/rpc/transport.py:73
↓ 1 callersMethodregister
Register without the common output args (--format/--output). The engine bench manages its own output lifecycle (CSV/JSON export, --qu
lmcache/cli/commands/bench/engine_bench/__init__.py:32
↓ 1 callersMethodregister
Register a periodic thread. Args: thread: The PeriodicThread to register
lmcache/v1/periodic_thread.py:436
↓ 1 callersMethodregister
Register a new instance-worker connection mapping. Args: msg: RegisterMsg from worker extra_config: Optional
lmcache/v1/cache_controller/controllers/registration_controller.py:102
↓ 1 callersMethodregister
Register *handler* for *kind*. ``payload_type`` is the msgspec decode type for the whole payload (e.g. ``list[ObjectKey]``).
lmcache/v1/multiprocess/native_completion.py:71
↓ 1 callersMethodregister
Subscribe all declared handlers to *bus*.
lmcache/v1/mp_observability/event_bus.py:85
↓ 1 callersFunctionregister_availability
Register an availability predicate for a device type. Args: device_type: The device type string (e.g., ``"cuda"``). predicate: A
lmcache/v1/platform/_registry.py:38
↓ 1 callersMethodregister_eviction_destination
Register an eviction destination for the eviction policy to use. Args: destination (EvictionDestination): The eviction d
lmcache/v1/distributed/eviction_policy/lru.py:61
↓ 1 callersMethodregister_fixed_buffers_from_allocator
Register allocator pages with io_uring when the allocator exposes them. Args: memory_allocator: Local CPU allocator that may expo
lmcache/v1/storage_backend/raw_block/core.py:447
↓ 1 callersMethodregister_l2_listener
Register a listener on all current and future L2 adapters. The listener is recorded so that adapters added later via :meth:`add_l2_ad
lmcache/v1/distributed/storage_manager.py:951
↓ 1 callersMethodregister_workers
Pre-register all workers before benchmark using DEALER-ROUTER mode
lmcache/tools/controller_benchmark/benchmark.py:267
↓ 1 callersMethodrelease_instance
Decrement the ref count; tear down the loop when it reaches 0.
lmcache/v1/multiprocess/mq.py:165
↓ 1 callersFunctionrelease_later
()
tests/v1/storage_backend/test_pd_backend_async.py:1078
↓ 1 callersMethodremote_xfer_handler_exists
(self, receiver_or_sender_id: str)
lmcache/v1/transfer_channel/nixl_channel.py:242
↓ 1 callersMethodremove
(self, key: CacheEngineKey, force: bool = True)
lmcache/v1/storage_backend/local_cpu_backend.py:271
↓ 1 callersMethodremove
(self, key, force=True)
lmcache/v1/storage_backend/remote_backend.py:594
↓ 1 callersMethodremove
remove a memory object. :param CacheEngineKey key: The key of the MemoryObj. :param bool force: Whether to it is a forced re
lmcache/v1/storage_backend/abstract_backend.py:223
↓ 1 callersMethodremove
( self, key: CacheEngineKey, force: bool = True, )
lmcache/v1/storage_backend/local_disk_backend.py:235
↓ 1 callersFunctionremoveInstance
(instanceId)
lmcache/v1/cache_controller/frontend/static/js/controller_app.js:417
↓ 1 callersMethodremove_adapter_state
Drop the eviction state for ``adapter_id``. Blocks until any in-progress eviction pass finishes (it holds the same lock), so the adap
lmcache/v1/distributed/storage_controllers/eviction_controller.py:245
↓ 1 callersFunctionremove_articles
(text)
benchmarks/rag/utils.py:164
↓ 1 callersMethodremove_chunks
Evict stale entries whose backing data is no longer in storage. Args: token_hashes: Token hashes of chunks to remove from the tab
lmcache/v1/multiprocess/modules/blend.py:242
↓ 1 callersMethodremove_chunks
Evict the given chunks from the matcher. Clears each chunk's table slot + poly hash so later probes cannot match it. Thread-safe.
lmcache/v1/multiprocess/modules/blend_v3.py:279
↓ 1 callersFunctionremove_pending_request
Remove a pending request from the dictionary.
examples/disagg_prefill_mp/disagg_proxy_server.py:223
↓ 1 callersFunctionremove_punc
(text)
benchmarks/rag/utils.py:170
← previousnext →4,501–4,600 of 13,371, ranked by callers