MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 4 callersMethod_complete_request
Store the retained-key bitmap and remove from in-flight tracking.
lmcache/v1/distributed/storage_controllers/prefetch_controller.py:1186
↓ 4 callersFunction_counter_callbacks
(meter: MagicMock)
tests/v1/mp_observability/subscribers/metrics/test_event_bus_self_metrics.py:32
↓ 4 callersMethod_create_key
( self, token_ids: list[int], start: int, end: int, request_id: str,
lmcache/integration/sglang/multi_process_adapter.py:196
↓ 4 callersMethod_ensure_peer_connection
( self, target_peer_init_url: str, force_update: bool = False, )
lmcache/v1/storage_backend/p2p_backend.py:497
↓ 4 callersFunction_escape_salt
Translate the empty-string salt to the URL sentinel.
lmcache/v1/multiprocess/http_apis/quota_api.py:41
↓ 4 callersFunction_events_of_type
(events: list, event_type: EventType)
tests/v1/distributed/test_distributed_storage_manager.py:720
↓ 4 callersMethod_file_path
(self, key: CacheEngineKey)
tests/v1/shm_allocator/shmfile_connector.py:202
↓ 4 callersMethod_format_safe_path
Generate a safe HTTP path for the S3 key. Flattens the key by replacing slashes with underscores and URL-encodes any special
lmcache/v1/storage_backend/connector/s3_connector.py:169
↓ 4 callersMethod_free_key
(self, key: CacheEngineKey)
lmcache/v1/hidden_state_store.py:355
↓ 4 callersFunction_gb
Convert bytes → GB for the JSON payload.
lmcache/v1/multiprocess/http_apis/quota_api.py:46
↓ 4 callersMethod_get_client
Get or create the per-thread GLIDE sync client. Creates a ``GlideClusterClient`` (cluster mode) or ``GlideClient`` (standalone mode)
lmcache/v1/storage_backend/connector/valkey_connector.py:137
↓ 4 callersMethod_get_committed_state_locked
(self, entry: _Entry)
lmcache/v1/storage_backend/dax/core.py:974
↓ 4 callersFunction_get_default_metadata
Get default metadata for testing. When *obj_size* is given the ``kv_shape`` is computed so that the resulting KV_2LTD tensor has exactly *obj
lmcache/v1/check/utils.py:59
↓ 4 callersFunction_get_head_size_view
Returns flattened views for index_copy/index_select. If engine_kv_format is provided, use it to interpret tensor layout explicitly. If n
lmcache/v1/gpu_connector/utils.py:628
↓ 4 callersMethod_get_next_task_id
Increment and return the next task ID. Must be called under _lock.
lmcache/v1/distributed/l2_adapters/native_connector_l2_adapter.py:385
↓ 4 callersFunction_get_ptr
(mv: Union[bytearray, memoryview])
lmcache/v1/storage_backend/connector/infinistore_connector.py:30
↓ 4 callersMethod_get_req_id
Extracts request ID from kwargs for logging.
lmcache/v1/cache_engine.py:277
↓ 4 callersMethod_get_request_tracker
(self, request_id: str)
lmcache/integration/vllm/lmcache_mp_connector.py:1314
↓ 4 callersMethod_get_request_tracker
(self, request_id: str)
lmcache/integration/vllm/lmcache_mp_connector_0201.py:1161
↓ 4 callersFunction_get_storage_manager
Extract storage_manager from the request app state. Returns: The StorageManager instance, or None if unavailable.
lmcache/v1/internal_api_server/vllm/backend_api.py:15
↓ 4 callersFunction_get_storage_manager
Resolve the shared ``StorageManager`` or return a 503 response. Returns either the ``StorageManager`` instance or a ``JSONResponse`` (503) re
lmcache/v1/multiprocess/http_apis/quota_api.py:51
↓ 4 callersFunction_get_strategies
()
lmcache/v1/lookup_client/record_strategies/__init__.py:34
↓ 4 callersMethod_gpu_meta
(self)
tests/cli/commands/bench/engine_bench/test_config.py:197
↓ 4 callersFunction_greedy_leftmost_dedupe
Drop overlapping matches keeping the leftmost — same as V2/V3 lookup.
tests/v1/multiprocess/test_optimized_lookup_v3.py:45
↓ 4 callersFunction_handle_exception
(operation: str, error: Exception)
lmcache/v1/internal_api_server/vllm/chunk_statistics_api.py:61
↓ 4 callersFunction_hex
(n: int, width: int = 4)
tests/v1/multiprocess/http_apis/test_l2_api.py:94
↓ 4 callersFunction_instances_response
(instances: list[dict])
tests/v1/multiprocess/test_p2p_controller.py:395
↓ 4 callersMethod_is_closed_or_closing
(self)
lmcache/v1/storage_backend/plugins/dax_backend.py:629
↓ 4 callersFunction_is_not_found_error
Return whether the exception represents a missing bucket or object.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:936
↓ 4 callersFunction_is_ptr_tensor
Return True when *x* is a 1-D pointer tensor (int64 or uint64).
lmcache/python_ops_fallback.py:806
↓ 4 callersMethod_key_string_to_object_path
Return the bucket object path for a serialized LMCache key.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:710
↓ 4 callersFunction_make_bitmap
(size: int)
lmcache/v1/distributed/l2_adapters/raw_block_l2_adapter.py:59
↓ 4 callersMethod_make_key_by_hash
( self, chunk_hash: int, request_configs: Optional[dict] = None )
lmcache/v1/token_database.py:233
↓ 4 callersFunction_make_metadata
Create a mock metadata with the given role.
tests/v1/internal_api_server/test_lookup_api.py:14
↓ 4 callersFunction_make_registry
(*instances: MPInstance)
tests/v1/mp_coordinator/test_eviction_manager.py:227
↓ 4 callersMethod_make_request
( self, method: str, key_str: str, *, body_stream=None, extra_headers=None )
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:834
↓ 4 callersFunction_normalize_url
Ensure *engine_url* has a scheme and ends with ``/v1``.
lmcache/cli/commands/bench/engine_bench/request_sender.py:23
↓ 4 callersMethod_offset_to_slot
Convert a data-slot byte offset to its slot index.
lmcache/v1/storage_backend/raw_block/core.py:1463
↓ 4 callersFunction_orig
(salt: str = "", object_group_id: int = 0)
tests/v1/distributed/serde/test_utils.py:9
↓ 4 callersFunction_parse_adapter
Helper: run the argparse plumbing on a single adapter JSON spec.
tests/v1/distributed/serde/test_factory.py:135
↓ 4 callersMethod_parse_eviction_config
Parse an optional ``"eviction"`` sub-dict from an adapter JSON spec. Expected format:: { "type": "mock"
lmcache/v1/distributed/l2_adapters/config.py:224
↓ 4 callersFunction_parse_mp
(argv: list[str])
tests/v1/multiprocess/test_config.py:105
↓ 4 callersFunction_parse_positive_int
(value: object, field_name: str)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:112
↓ 4 callersFunction_patched_pd
()
tests/v1/storage_backend/test_pd_backend_async.py:155
↓ 4 callersFunction_pd_backend_patches
()
tests/v1/storage_backend/test_pd_backend_async.py:137
↓ 4 callersMethod_poll_coordinator_match
Poll the coordinator match result, deferring until it resolves. Mirrors the prefix/sparse legs: ``return None`` to defer while pending.
lmcache/v1/multiprocess/modules/blend_v3.py:1267
↓ 4 callersMethod_prepare_write_payload
Prepare the payload buffer and lengths for a raw-block write. Args: memory_obj: Source object to persist. Returns:
lmcache/v1/storage_backend/raw_block/core.py:1070
↓ 4 callersMethod_query
(self, request_data: dict[str, Any], *, chat: bool)
lmcache/cli/commands/query/_request.py:336
↓ 4 callersFunction_quota_manager
Return the shared :class:`QuotaManager` from ``app.state``.
lmcache/v1/mp_coordinator/http_apis/l2_api.py:44
↓ 4 callersFunction_range
(prefix: str, tokens: list[int])
tests/v1/mp_coordinator/test_blend_directory_api.py:21
↓ 4 callersFunction_registry
Return the shared instance registry from app state. Args: request: The incoming request. Returns: The shared :class:`Instanc
lmcache/v1/mp_coordinator/http_apis/instances_api.py:33
↓ 4 callersMethod_release_lease
Release a lease to free server resources immediately. Args: key: The cache key lease_id: The lease ID to rel
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:299
↓ 4 callersFunction_resolve_extra_config
Resolve extra_config against :class:`ExtraConfigDefault`. Keys in *extra_config* are expected to carry the ``lmcache.mp.`` prefix (e.g. ``lmc
lmcache/integration/vllm/vllm_multi_process_adapter.py:79
↓ 4 callersFunction_sample_labels
Return labels as Prometheus exposes them in collected samples.
tests/test_observability.py:313
↓ 4 callersMethod_set_healthy
Set the health status
lmcache/v1/health_monitor/base.py:271
↓ 4 callersMethod_shmfs_cfg
Read extra_config['shmfs.<key>'].
tests/v1/shm_allocator/shmfile_connector.py:121
↓ 4 callersFunction_simulate_lru
Replay *access_seq* against an LRU of *capacity*; return (hits, misses). Args: access_seq: Ordered list of cache keys to access.
tests/cli/commands/bench/engine_bench/workloads/test_prefix_suffix_tuner.py:571
↓ 4 callersFunction_store_range
(prefix: str, tokens: list[int])
tests/v1/mp_coordinator/test_blend_client.py:76
↓ 4 callersFunction_strip_leading_slash
Normalise a name to the bare form (no leading ``/``). Callers historically embed the POSIX leading slash; we keep the on-wire name slash-pref
lmcache/v1/multiprocess/posix_shm.py:48
↓ 4 callersFunction_sum
(name: str)
tests/v1/mp_observability/subscribers/metrics/test_l0_l1_throughput.py:93
↓ 4 callersFunction_tick
(triggered: bool, usage: float = 0.5)
tests/v1/mp_observability/subscribers/metrics/test_l1_eviction_loop.py:48
↓ 4 callersFunction_unavailable_response
Return a 503 response when the engine is unavailable.
lmcache/v1/internal_api_server/vllm/backend_api.py:31
↓ 4 callersMethod_validate_uring_cmd_chunk
Validate one NVMe raw-command transfer range. Args: offset: Device byte offset for the transfer. total_len: Transfer
lmcache/v1/storage_backend/raw_block/core.py:1112
↓ 4 callersFunction_wait_match
(client: BlendCoordinatorClient, rid: str, timeout: float = 2.0)
tests/v1/mp_coordinator/test_blend_client.py:81
↓ 4 callersMethod_writable_devices_by_usage_locked
( self, *, exclude_device_id: Optional[int], )
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1203
↓ 4 callersMethod_write_keys_and_wait_for_l2
Write keys to L1 via StorageManager and wait for L2 store.
tests/v1/distributed/test_distributed_storage_manager.py:539
↓ 4 callersMethodadd_list_section
Add a section that belongs to a list group. In terminal output, renders as a normal section with *label* as the header. In JSON outpu
lmcache/cli/metrics/metrics.py:91
↓ 4 callersFunctionadd_observability_args
Add observability configuration arguments to an existing parser. Args: parser: The argument parser to add arguments to. Returns:
lmcache/v1/mp_observability/config.py:92
↓ 4 callersFunctionadd_pending_module
Register a module path for deferred import. The module will only be imported when a type name lookup misses the eager registry. Importing th
lmcache/v1/distributed/l2_adapters/factory.py:84
↓ 4 callersMethodallocate
( self, shapes: Union[torch.Size, list[torch.Size]], dtypes: Union[torch.dtype, list[t
lmcache/v1/storage_backend/nixl_storage_backend.py:870
↓ 4 callersFunctionassert_contiguous
Assert that *tensor* has a contiguous physical layout with zero offset. LMCache transfer kernels assume logical and physical views match for
lmcache/v1/gpu_connector/utils.py:39
↓ 4 callersMethodasync_release_reservation
Release reservation (asyncio variant). Notifies all waiters. :param req_id: The request identifier whose reservation to release.
lmcache/v1/storage_backend/pd_backend_async.py:182
↓ 4 callersMethodbatched_free
Return a batch of storage slot indices back to the pool. Args: obj_indices: Indices previously obtained from ``batched_a
lmcache/v1/distributed/l2_adapters/nixl_store_l2_adapter.py:115
↓ 4 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:1303
↓ 4 callersMethodbatched_from_gpu
(self, memory_objs, starts, ends, **kwargs)
lmcache/v1/gpu_connector/gpu_connectors.py:1632
↓ 4 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:1111
↓ 4 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:1652
↓ 4 callersMethodbatched_get_blocking
Restore a batch of DAX-backed cache entries synchronously. The returned list preserves the input order. Entries that are missing or n
lmcache/v1/storage_backend/plugins/dax_backend.py:516
↓ 4 callersMethodbatched_submit_put_task
Submit batched put tasks to transfer KV caches to peer. The NIXL transfer is offloaded to a background thread to avoid blocking
lmcache/v1/storage_backend/pd_backend.py:496
↓ 4 callersMethodbatched_to_gpu
(self, memory_objs, starts, ends, **kwargs)
lmcache/v1/gpu_connector/gpu_connectors.py:1627
↓ 4 callersMethodbatched_to_gpu
This function is a generator that moves the KV cache from the memory objects to paged GPU memory. The first iteration will prepare so
lmcache/v1/gpu_connector/xpu_connectors.py:1524
↓ 4 callersMethodcalculate_slots
Calculate the number of slots for the specified number of tokens. Assuming the number of tokens are already aligned.
lmcache/v1/kv_layer_groups.py:225
↓ 4 callersFunctioncall_block_kernel
( vllm_tensors: list[torch.Tensor], mem_objects: list[torch.Tensor], block_ids: list[int], eng
tests/v1/test_mp_mem_kernels.py:207
↓ 4 callersFunctioncausal_prefill_attention
Standard causal prefill attention via Triton. Args: query: [M, H, D] key: [M, H, D] value: [M, H, D] sm_scale: 1/
lmcache/v1/compute/attention/triton_kernels/block_sparse_attention.py:463
↓ 4 callersMethodcheck
Perform a health check on remote backend, which includes the following checks: - get_blocking/batched_get_blocking, if failed count
lmcache/v1/health_monitor/checks/remote_backend_check.py:156
↓ 4 callersMethodclear
Clear data in the storage manager. Args: force: If True, clear ALL objects including locked ones. This m
lmcache/v1/distributed/storage_manager.py:903
↓ 4 callersMethodclose
(self)
tests/v1/distributed/transfer_channel/test_factory.py:57
↓ 4 callersMethodclose
Close socket and context
tests/v1/storage_backend/test_p2p_backend_with_controller.py:118
↓ 4 callersMethodclose
Close the underlying HTTP client.
lmcache/cli/commands/bench/engine_bench/request_sender.py:212
↓ 4 callersMethodclose
(self)
lmcache/v1/distributed/l2_adapters/nixl_store_l2_adapter.py:411
↓ 4 callersMethodclose
(self)
lmcache/v1/distributed/l2_adapters/nixl_store_l2_adapter.py:581
↓ 4 callersMethodclose
Stop worker pools and release DAX resources. The call waits for already submitted worker tasks to finish before closing mapped DAX co
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:590
↓ 4 callersMethodclose
(self)
lmcache/v1/multiprocess/transfer_context/shm.py:211
↓ 4 callersMethodclose
Deregister the handle and close the file.
lmcache/v1/storage_backend/hipfile_shim.py:92
↓ 4 callersMethodclose
Close the storage backend.
lmcache/v1/storage_backend/pd_backend_async.py:1620
↓ 4 callersMethodclose
(self)
lmcache/v1/storage_backend/connector/redis_connector.py:232
↓ 4 callersFunctioncompute_kv_layout
Compute KV layout metadata from KV tensors. Args: kv_caches: Per-layer KV tensor mapping. layout_hints: Optional engine layout hi
lmcache/v1/multiprocess/transfer_context/base.py:263
↓ 4 callersFunctioncontainerArgs
containerArgs returns the args of the first (and only) container in the DaemonSet pod template — the smoke contract requires exactly one.
operator/test/e2e/smoke_helpers_test.go:211
↓ 4 callersMethodcontains
(self, key: Any)
lmcache/tools/cache_simulator/lru_cache.py:89
↓ 4 callersFunctioncountLMCacheStored
countLMCacheStored runs `kubectl logs` on the LMCache MP server pod and returns the number of matching "Stored N tokens" lines seen so far. Returns 0
operator/test/e2e/vllm_integration_smoke_test.go:244
← previousnext →1,401–1,500 of 13,371, ranked by callers