MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 4 callersFunctionparse_args
()
lmcache/v1/basic_check.py:12
↓ 4 callersFunctionparse_args_to_mp_server_config
Convert parsed command line arguments to an MPServerConfig. Args: args: Parsed arguments from the argument parser. Returns:
lmcache/v1/multiprocess/config.py:368
↓ 4 callersFunctionparse_args_to_observability_config
Convert parsed command line arguments to an ObservabilityConfig. Args: args: Parsed arguments from the argument parser. Returns:
lmcache/v1/mp_observability/config.py:227
↓ 4 callersFunctionparse_hfbucket_handle
Parse a Hugging Face bucket handle into bucket id and prefix. Args: bucket_handle: Handle in ``hf://buckets/<namespace>/<bucket>/<prefix>
lmcache/v1/storage_backend/connector/hfbucket_connector.py:729
↓ 4 callersMethodpin
(self, key: CacheEngineKey)
lmcache/v1/storage_backend/local_cpu_backend.py:255
↓ 4 callersMethodpin
( self, key: CacheEngineKey, )
lmcache/v1/storage_backend/local_disk_backend.py:213
↓ 4 callersFunctionping
GET *url* and return ``(status, rtt_ms, error_msg)``. Returns: ``("OK", rtt_ms, None)`` on HTTP 200. ``("FAIL", rtt_ms, detail)``
lmcache/cli/commands/ping.py:45
↓ 4 callersMethodping
Respond to a ping and refresh the sender's liveness. Args: instance_id: The sender's worker instance ID, or None for an
lmcache/v1/multiprocess/modules/management.py:128
↓ 4 callersMethodpost_blocking
(self, handle: NixlXferHandle)
lmcache/v1/storage_backend/nixl_storage_backend.py:525
↓ 4 callersMethodpost_init
Post initialization of the KV controller.
lmcache/v1/cache_controller/controllers/kv_controller.py:97
↓ 4 callersFunctionprint_output
( llm: LLM, prompt: list[int], sampling_params: SamplingParams, req_str: str, )
examples/blend_kv_v1/blend.py:85
↓ 4 callersFunctionpullSecretNames
pullSecretNames returns the names in the pod's imagePullSecrets.
operator/internal/webhook/pod_injector_test.go:190
↓ 4 callersMethodpush
(self)
lmcache/v1/cache_controller/worker.py:517
↓ 4 callersMethodput_msg
Put a message into the message queue.
lmcache/v1/cache_controller/worker.py:378
↓ 4 callersMethodquery_lookup_and_lock_result
(self, task_id: L2TaskId)
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:192
↓ 4 callersFunctionread_counters
Snapshot counter values, summed across all attribute combinations. Accumulates into the result dict (rather than overwriting) when the same m
tests/v1/mp_observability/subscribers/metrics/otel_setup.py:30
↓ 4 callersFunctionread_tagged_counters
Snapshot all counter values keyed by (metric_name, attrs). Works for tagged counters where multiple data points differ only by their attribut
tests/v1/mp_observability/subscribers/metrics/counter_helpers.py:31
↓ 4 callersMethodreconcileConnectionConfigMap
reconcileConnectionConfigMap creates or updates the connection ConfigMap.
operator/internal/controller/reconcile_helpers.go:217
↓ 4 callersMethodreconcileDaemonSet
reconcileDaemonSet creates or updates the DaemonSet.
operator/internal/controller/reconcile_helpers.go:120
↓ 4 callersMethodreconcileLookupService
reconcileLookupService creates or updates the node-local lookup Service.
operator/internal/controller/reconcile_helpers.go:162
↓ 4 callersMethodreconcileMetricsService
reconcileMetricsService creates or updates the headless metrics Service.
operator/internal/controller/reconcile_helpers.go:190
↓ 4 callersMethodrecord_answer
Record a completed exchange and mark session as ready.
lmcache/cli/commands/bench/engine_bench/workloads/multi_round_chat.py:142
↓ 4 callersFunctionrecover_gpu_connector_states
(gpu_connector)
tests/v1/utils.py:162
↓ 4 callersFunctionregister
Register an MP server with the coordinator and return its id. Args: client: The HTTP client to send with. base_url: Coordinator b
lmcache/v1/mp_coordinator/registrar.py:33
↓ 4 callersMethodregister
Register this command with the CLI argument parser. This method is not typically overridden. It calls :meth:`name`, :meth:`help`, an
lmcache/cli/commands/base.py:78
↓ 4 callersMethodregistered_qualnames
Return the list of currently registered qualnames. Returns: A new list of qualname strings, suitable for inspection or
lmcache/cli/commands/trace/_dispatch.py:102
↓ 4 callersMethodrelease
Return *slots* to the semaphore and wake waiting acquirers. Args: slots: Number of slots to release. No-op if <= 0.
examples/disagg_prefill/disagg_proxy_server.py:63
↓ 4 callersMethodrelease_read
Release a read lock.
lmcache/v1/cache_controller/locks.py:52
↓ 4 callersMethodremove
Remove the key from the storage backend. :param key: The key to remove.
lmcache/v1/storage_backend/pd_backend_async.py:1492
↓ 4 callersMethodrename
(old_fname: Path, new_fname: Path)
lmcache/v1/storage_backend/connector/hf3fs_connector.py:607
↓ 4 callersMethodreport_status
Return a health and capacity snapshot for this adapter. Returns: Dictionary containing health, DAX capacity, slot occupancy,
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:632
↓ 4 callersFunctionretrieve_keys
Retrieve keys one at a time using the single-key API.
tests/v1/multiprocess/test_cache_server.py:206
↓ 4 callersMethodretrieve_layer
Retrieve the KV cache in a layerwise manner. :param torch.Tensor tokens: The tokens of the corresponding KV caches. :param
lmcache/v1/cache_engine.py:969
↓ 4 callersFunctionround_down
Round down x to the nearest multiple of y.
lmcache/utils.py:103
↓ 4 callersFunctionrun_perf_test_with_timeout
Common performance test framework with timeout handling
lmcache/v1/check/utils.py:234
↓ 4 callersMethodseek
(self, offset, whence=0)
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:281
↓ 4 callersMethodselect_l1_retentions
Retain all prefetched keys permanently in L1.
lmcache/v1/distributed/storage_controllers/prefetch_policy.py:184
↓ 4 callersMethodsend_cmd
(self, cmd: str)
tests/v1/shm_allocator/verify_shmfile_tcp.py:121
↓ 4 callersMethodsend_heartbeat
Send heartbeat request
lmcache/lmcache_frontend/heartbeat.py:93
↓ 4 callersMethodsend_messages
Send multiple messages via ZMQ PUSH socket Args: messages: List of messages to send Returns: Time taken to s
lmcache/tools/controller_benchmark/benchmark.py:196
↓ 4 callersFunctionsend_request_to_service
Send a request to a service using a persistent client.
examples/disagg_prefill/disagg_proxy_server.py:399
↓ 4 callersMethodserialize
(self, src, dst)
tests/v1/distributed/serde/test_multi.py:364
↓ 4 callersMethodset
(self, key, value)
tests/conftest.py:180
↓ 4 callersMethodset_app_config
Set application configuration for heartbeat reporting
lmcache/lmcache_frontend/heartbeat.py:87
↓ 4 callersFunctionset_shape_desc_dtype
Best-effort ``shape_desc.dtype = dtype``. The pure-Python ``PageBufferShapeDesc`` exposes a ``dtype`` slot so the CPU fallback kernel can dis
lmcache/python_ops_fallback.py:348
↓ 4 callersMethodsetup
Setup ZMQ sockets
lmcache/tools/controller_benchmark/benchmark.py:103
↓ 4 callersFunctionshm_map_readwrite
Open an existing shared-memory segment and return its address. ``nbytes`` must match the segment's actual size; ``mmap`` will raise on a mism
lmcache/v1/multiprocess/posix_shm.py:148
↓ 4 callersMethodshould_request_full_sync
Check if a worker should perform full sync. Returns: Tuple of (need_sync, reason)
lmcache/v1/cache_controller/controllers/full_sync_tracker.py:82
↓ 4 callersFunctionsnapshot
Capture counters before the test; yield a callable that returns deltas.
tests/v1/mp_observability/subscribers/metrics/test_l1_eviction_loop.py:38
↓ 4 callersMethodstart
Start the storage controller. This function should be implemented by subclasses to start any necessary threads or processes.
lmcache/v1/distributed/storage_controller.py:15
↓ 4 callersMethodstart_ret
Create FullSyncStartRetMsg.
tests/v1/cache_controller/conftest.py:197
↓ 4 callersMethodstart_services
Start all managed services. Managed services include: - InternalAPIServer: HTTP server exposing internal APIs for
lmcache/v1/manager.py:165
↓ 4 callersMethodstatus_ret
Create FullSyncStatusRetMsg.
tests/v1/cache_controller/conftest.py:204
↓ 4 callersMethodstop
(self)
tests/cli/commands/bench/test_server_bench.py:443
↓ 4 callersMethodstop
Stop the storage controller. This function should be implemented by subclasses to stop any running threads or processes.
lmcache/v1/distributed/storage_controller.py:24
↓ 4 callersMethodsubmit_lookup_and_lock_task
( self, keys: list[ObjectKey], layout_desc: MemoryLayoutDesc )
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:183
↓ 4 callersMethodsubmit_set
Submit a SET operation.
lmcache/v1/storage_backend/connector/valkey_connector.py:254
↓ 4 callersMethodsubmit_unlock
(self, keys: list[ObjectKey])
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:196
↓ 4 callersMethodsubmit_unlock
(self, keys: list[ObjectKey])
lmcache/v1/distributed/l2_adapters/nixl_store_dynamic_l2_adapter.py:437
↓ 4 callersMethodtell
(self)
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:290
↓ 4 callersMethodtime_to_store
(self)
lmcache/observability.py:198
↓ 4 callersFunctiontimed_generate
(llm, prompt, sampling_params, label)
tests/v1/compute/attention/test_triton_sparse_e2e.py:74
↓ 4 callersMethodto_gpu
Expect a kwarg 'kvcaches' which is a nested tuple of K and V tensors. The kvcaches should correspond to the "WHOLE token sequence". N
lmcache/v1/gpu_connector/xpu_connectors.py:1317
↓ 4 callersMethodtotal_duration_s
Return replay wall-clock duration in seconds. Returns: ``mark_end - mark_start`` if both were set, else 0.0.
lmcache/cli/commands/trace/_stats.py:160
↓ 4 callersMethodupdate_interval_remote_time_to_get
(self, get_time: float)
lmcache/observability.py:541
↓ 4 callersMethodupdate_interval_remote_time_to_get_sync
(self, get_time_sync: float)
lmcache/observability.py:549
↓ 4 callersMethodupdate_local_storage_usage
(self, usage: int)
lmcache/observability.py:527
↓ 4 callersMethodupdate_on_force_evict
( self, key: KeyType, )
lmcache/v1/storage_backend/cache_policy/lfu.py:67
↓ 4 callersMethodupdate_remote_ping_latency
(self, latency: float)
lmcache/observability.py:553
↓ 4 callersFunctionvalidate_node
Validate a single node configuration dict. Args: node: Candidate node dict. is_proxy: Reserved for future proxy-specific validati
lmcache/lmcache_frontend/app.py:303
↓ 4 callersFunctionvalidate_raw_block_io_options
Validate numeric raw-block I/O engine options. Args: iouring_queue_depth: Queue depth for the Rust io_uring path. Raises: Va
lmcache/v1/storage_backend/raw_block/core.py:94
↓ 4 callersMethodwait_for_completion
Wait for async queue to be processed. Args: timeout: Maximum time to wait in seconds Returns: True if queue
lmcache/v1/lookup_client/record_strategies/base.py:280
↓ 4 callersFunctionwait_for_condition
Poll until a predicate returns True or timeout.
tests/v1/distributed/test_distributed_storage_manager.py:144
↓ 4 callersFunctionwait_for_condition
(predicate, timeout: float = 5.0, interval: float = 0.02)
tests/v1/distributed/test_dax_l2_adapter.py:108
↓ 3 callersFunctionBuildCBConnectionConfigMap
BuildCBConnectionConfigMap creates the <engine>-connection ConfigMap carrying the CBKVConnector kv-transfer-config JSON (design §7). The JSON points v
operator/internal/resources/cacheblend_engine.go:144
↓ 3 callersFunctionBuildCoordinatorArgs
BuildCoordinatorArgs maps the LMCacheCoordinatorSpec config fields to the `lmcache coordinator` CLI flags. User-supplied extraArgs are appended last s
operator/internal/resources/coordinator_deployment.go:93
↓ 3 callersMethodComputeKVRank
Compute the kv_rank from world_size and worker_id Args: world_size (int): The total number of workers (include TP + PP)
lmcache/v1/distributed/api.py:121
↓ 3 callersFunctionCreateSerde
( serde_type: str, metadata: LMCacheMetadata, config: LMCacheEngineConfig, )
lmcache/v1/storage_backend/naive_serde/__init__.py:21
↓ 3 callersFunctionCreateTransferChannel
Create a transfer channel based on the specified channel type. Supports "nixl" and "mock_memory" channel types. If nixl is not available,
lmcache/v1/transfer_channel/__init__.py:10
↓ 3 callersMethodDeepCopy
DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new LMCacheEngine.
operator/api/v1alpha1/zz_generated.deepcopy.go:726
↓ 3 callersMethodGetInstanceOrNone
Returns the singleton instance of CacheControllerPrometheusLogger if it exists, otherwise returns None.
lmcache/v1/cache_controller/observability.py:154
↓ 3 callersMethodGetOrCreate
()
lmcache/observability.py:832
↓ 3 callersMethodSetupWithManager
SetupWithManager sets up the controller with the Manager.
operator/internal/controller/lmcacheengine_controller.go:124
↓ 3 callersMethod__init__
( self, host: str, port: int, loop: asyncio.AbstractEventLoop, local_c
lmcache/v1/storage_backend/connector/redis_connector.py:43
↓ 3 callersMethod__init__
(self)
lmcache/v1/storage_backend/connector/redis_adapter.py:21
↓ 3 callersMethod_abort_request
Clean up request state and notify receiver of cancellation. Sends CancelNotif to the receiver so it can release allocated keys and re
lmcache/v1/storage_backend/pd_backend_async.py:1112
↓ 3 callersMethod_alloc_id
(self)
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:260
↓ 3 callersFunction_api_error_response
(exc: L2ReconfigureError)
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:142
↓ 3 callersMethod_append_free_slot_locked
Add a slot to the free list while ``self._lock`` is held.
lmcache/v1/storage_backend/raw_block/core.py:1478
↓ 3 callersFunction_async_readinto_full
Async version of :func:`_readinto_full`.
lmcache/v1/distributed/l2_adapters/fs_l2_adapter.py:83
↓ 3 callersMethod_async_save_bytes_to_disk
Convert KV to bytes and async store bytes to disk. :param on_complete_callback: Optional callback invoked after the GDS
lmcache/v1/storage_backend/gds_backend.py:603
↓ 3 callersFunction_attrs_of_nonzero_dps
(name: str)
tests/v1/mp_observability/subscribers/metrics/test_l0_l1_throughput.py:98
↓ 3 callersFunction_attrs_of_nonzero_dps
(name: str)
tests/v1/mp_observability/subscribers/metrics/test_l2_throughput.py:146
↓ 3 callersFunction_bm
(num_keys: int, set_indices: list[int])
tests/v1/distributed/test_bitmap_ops.py:233
↓ 3 callersMethod_bm
(n, idxs)
tests/v1/distributed/test_prefetch_controller.py:1262
↓ 3 callersFunction_build
Build a synthetic normalized kv_caches for the named format.
tests/v1/gpu_connector/test_kv_format_specs.py:33
↓ 3 callersFunction_build_engine
( *, name: str, cfg, connector, num_layers: int, chunk_size: int, num_heads: int,
tests/benchmarks/test_xpu_connector_benchmark.py:188
↓ 3 callersMethod_byte_view
Return a byte-addressable memoryview over a Python buffer. Args: buf: Object implementing the Python buffer protocol. Re
lmcache/v1/storage_backend/raw_block/core.py:982
↓ 3 callersMethod_cache_discard
(self, chunk_hash: int)
lmcache/v1/storage_backend/nixl_storage_backend.py:1462
↓ 3 callersMethod_check_timeouts
Check all registered pinned objects for timeout. Returns: tuple: (pinned_count, timeout_count, force_unpin_success_count)
lmcache/v1/pin_monitor.py:120
← previousnext →1,601–1,700 of 13,371, ranked by callers