Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/LMCache/LMCache
/ functions
Functions
13,371 in github.com/LMCache/LMCache
⨍
Functions
13,371
◇
Types & classes
2,044
↳
Endpoints
181
↓ 4 callers
Method
counts
(self)
tests/v1/distributed/test_s3_l2_adapter.py:83
↓ 4 callers
Function
cpu_allocator
Create and cleanup CPU memory allocator
tests/v1/storage_backend/test_p2p_backend_with_controller.py:269
↓ 4 callers
Function
create_full_chunk_memory_obj
Allocate and initialize a full chunk memory object for upload tests.
tests/v1/storage_backend/test_hfbucket_connector.py:209
↓ 4 callers
Function
create_key
(chunk_hash: str)
tests/v1/test_nixl_storage.py:81
↓ 4 callers
Function
create_nixl_cpu_config
Create a config with NIXL CPU mode enabled.
tests/v1/storage_backend/test_local_cpu_backend_nixl_paged.py:37
↓ 4 callers
Method
create_stats_client
(self, **kwargs)
tests/v1/lookup_client/test_chunk_statistics_lookup_client.py:63
↓ 4 callers
Function
create_test_config
Create a test configuration.
tests/v1/cache_controller/conftest.py:37
↓ 4 callers
Function
create_test_config
(extra_overrides=None)
tests/v1/storage_backend/test_bigtable_integration.py:38
↓ 4 callers
Function
create_test_config
Create a test configuration for storage plugin testing.
tests/v1/storage_backend/test_storage_plugin.py:149
↓ 4 callers
Function
create_test_config
Create a test configuration for FSConnector.
tests/v1/storage_backend/test_fs_connector.py:22
↓ 4 callers
Function
create_test_key
Create a test CacheEngineKey.
tests/v1/storage_backend/test_storage_plugin.py:138
↓ 4 callers
Function
create_test_metadata
Create test metadata for testing.
tests/v1/storage_backend/test_storage_plugin.py:167
↓ 4 callers
Method
debug_get_locked_key_count
Return the number of currently locked keys. This method is intended for testing and debugging only. Returns: in
lmcache/v1/distributed/l2_adapters/mock_l2_adapter.py:304
↓ 4 callers
Function
decode_object_key
Deserialize an ObjectKey from raw-block's reversible encoding. Args: encoded: Encoded key string produced by ``object_key_to_string``.
lmcache/v1/storage_backend/raw_block/key_codec.py:53
↓ 4 callers
Method
decrease_pin_count
(self)
lmcache/v1/distributed/l2_adapters/nixl_store_l2_adapter.py:66
↓ 4 callers
Method
deregister
De-register the lmcache worker from the controller.
lmcache/v1/cache_controller/worker.py:243
↓ 4 callers
Method
deregister_gpu_buffer
Reverse of :meth:`register_gpu_buffer`: deregister its regions + stream. Args: buffer: The buffer passed to :meth:`register_gpu_b
lmcache/v1/gpu_connector/gds_context.py:149
↓ 4 callers
Function
do_dequantize
t: [nlayers, ntokens, nchannels] bins: [nlayers] maxtensors: [nlayers, ntokens, 1]
lmcache/storage_backend/serde/cachegen_decoder.py:33
↓ 4 callers
Method
do_shuffle
(self, t)
lmcache/v1/compute/positional_encoding.py:26
↓ 4 callers
Method
end_msg
Create FullSyncEndMsg.
tests/v1/cache_controller/conftest.py:187
↓ 4 callers
Function
filterProxies
()
lmcache/lmcache_frontend/static/js/app.js:140
↓ 4 callers
Method
format
(self, record)
lmcache/logging.py:33
↓ 4 callers
Function
formatTokenCountStr
(count)
lmcache/lmcache_frontend/static/js/app.js:700
↓ 4 callers
Method
free
Return the slab regions of the given objects to the address manager. Args: mem_objs: Objects to free (slab-anchored :class:`GDSMe
lmcache/v1/distributed/memory_manager/gds_l1_memory_manager.py:94
↓ 4 callers
Method
free_lookup_locks
Release read locks acquired during lookup. Hashes are computed only for chunks in ``[start, end)`` to avoid unnecessary work on token
lmcache/v1/multiprocess/modules/lookup.py:414
↓ 4 callers
Method
freeze
Set the freeze mode for the cache engine. When freeze mode is enabled: - All store operations will be skipped (no new data s
lmcache/v1/cache_engine.py:332
↓ 4 callers
Method
from_dict
(cls, d: dict)
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:73
↓ 4 callers
Method
from_dict
(d)
lmcache/v1/memory_management.py:168
↓ 4 callers
Method
from_gpu
Expect a kwarg 'kvcaches' which is a nested tuple of K and V tensors. The kvcaches should correspond to the "WHOLE token sequence". W
lmcache/v1/gpu_connector/xpu_connectors.py:1372
↓ 4 callers
Method
from_ipc_handle
Reconstruct a StubEvent from an IPC handle. Args: device: The device to associate with the event. handle: The IPC han
lmcache/v1/platform/cpu/stub_cpu_device.py:98
↓ 4 callers
Method
from_json
Load from a saved config JSON dict.
lmcache/cli/commands/bench/engine_bench/interactive/state.py:269
↓ 4 callers
Method
gather
(self, values: Sequence[_T])
tests/v1/storage_backend/raw_block_test_utils.py:226
↓ 4 callers
Method
get
Read a file into a newly allocated memory_obj via the worker subprocess.
tests/v1/shm_allocator/shmfile_connector.py:232
↓ 4 callers
Method
get_all_results
Return all raw results for CSV export.
lmcache/cli/commands/bench/engine_bench/stats.py:210
↓ 4 callers
Method
get_async_response
( self, url: str, *, timeout: Optional[float] = None, extra_headers: O
lmcache/connections.py:73
↓ 4 callers
Function
get_attention_backend
Return a representative attention-backend label for the format. Diagnostic only. A format may be produced by several (engine, attention-backe
lmcache/v1/gpu_connector/utils.py:122
↓ 4 callers
Method
get_bit_set
(self)
lmcache/v1/utils/bloom_filter.py:92
↓ 4 callers
Method
get_cache_size_per_token
Returns the cache size per token (in bytes), summed across all kernel groups.
lmcache/v1/platform/cuda/cache_context.py:255
↓ 4 callers
Method
get_context
Return the OTel context for an open span, or ``None``. Convenience wrapper over :meth:`get` for callers that only need the context to
lmcache/v1/mp_observability/subscribers/tracing/span_registry.py:75
↓ 4 callers
Method
get_count_by_level
Get counts of threads for a specific level. Returns: Dict with keys: total, running, active
lmcache/v1/periodic_thread.py:499
↓ 4 callers
Method
get_current_retrieve_stats
(self)
lmcache/observability.py:326
↓ 4 callers
Method
get_device_properties
Return device properties for the given device. Args: device: The device index or identifier to query. Ignored in
lmcache/v1/platform/cpu/stub_cpu_device.py:315
↓ 4 callers
Function
get_engine_kv_shape_description
Return a human-readable symbolic shape legend for the Engine KV format. Uses short names matching the ``EngineKVFormat`` enum convention: NB=
lmcache/v1/gpu_connector/utils.py:109
↓ 4 callers
Function
get_env_name
(attr_name: str)
lmcache/v1/config_base.py:263
↓ 4 callers
Function
get_expected_count
Calculate expected token count based on save_unfull_chunk setting. Args: token_len: Total token length save_unfull_chunk: Whether
tests/benchmarks/test_benchmark.py:37
↓ 4 callers
Method
get_global_progress
Get the global sync progress. Returns: Progress as a float between 0.0 and 1.0 Note: Uses cached worker list fo
lmcache/v1/cache_controller/controllers/full_sync_tracker.py:320
↓ 4 callers
Method
get_kernel_group_shape_dtype
( self, num_tokens: int, kernel_group_idx: int, )
tests/v1/platform/test_cache_context_dispatch.py:108
↓ 4 callers
Function
get_l1_object_count
Return current L1 object count via public report_status.
tests/v1/distributed/serde/test_serde_e2e.py:193
↓ 4 callers
Method
get_memory_usage
Get the current memory usage. This function will mainly be used to support eviction decision. Returns: tuple[int
lmcache/v1/distributed/memory_manager/l1_memory_manager.py:168
↓ 4 callers
Method
get_memory_usage_bytes
(self)
lmcache/v1/utils/bloom_filter.py:89
↓ 4 callers
Method
get_missing_required
Return phase-1 items that still need user input. Includes both required items that are unset, and non-required phase-1 items (like ``
lmcache/cli/commands/bench/engine_bench/interactive/state.py:96
↓ 4 callers
Function
get_mock_backend
Get the mock storage plugin instance from storage manager or backends dict. Handles both StorageManager and OrderedDict[str, StorageBackendIn
tests/v1/storage_backend/test_storage_plugin.py:186
↓ 4 callers
Method
get_mode
Get registered mode function. Returns None if the mode is not found.
lmcache/v1/check/__init__.py:55
↓ 4 callers
Method
get_peer_init_url
Get the URL for a given instance and worker ID.
lmcache/v1/cache_controller/controllers/registration_controller.py:71
↓ 4 callers
Method
get_response
( self, url: str, *, stream: bool = False, timeout: Optional[float] =
lmcache/connections.py:53
↓ 4 callers
Method
get_shape
(self, num_tokens: int)
lmcache/v1/gpu_connector/xpu_connectors.py:1437
↓ 4 callers
Method
get_status
Get the current status of the thread. Returns: Dict: Thread status information
lmcache/v1/periodic_thread.py:364
↓ 4 callers
Method
get_sync_client
(self)
lmcache/connections.py:28
↓ 4 callers
Method
handle_full_sync_status
Handle full sync status query from a worker. Returns the sync status including any missing batches that need resending.
lmcache/v1/cache_controller/controllers/kv_controller.py:345
↓ 4 callers
Method
has
Return ``True`` if a handler is registered for *qualname*.
lmcache/cli/commands/trace/_dispatch.py:98
↓ 4 callers
Function
hasEnvAll
hasEnvAll reports whether envs contains an env var named name set to the literal "all" (the value the GPU passthrough vars are always set to).
operator/internal/resources/resources_test.go:1141
↓ 4 callers
Function
http_request
Send an HTTP request and return the parsed JSON response. Args: method: HTTP method (GET, POST, PUT, DELETE). url: Full URL to re
lmcache/cli/commands/quota/_helpers.py:38
↓ 4 callers
Method
inflight_io_count
Return the number of active raw-device I/O operations.
lmcache/v1/storage_backend/plugins/rust_raw_block_backend.py:217
↓ 4 callers
Function
init_observability
Initialize OTel providers, EventBus, and register subscribers. This is the single entry-point that every MP server calls at startup. Returns
lmcache/v1/mp_observability/config.py:266
↓ 4 callers
Function
init_otel_metrics
Set up the OpenTelemetry MeterProvider. Args: otlp_endpoint: OTLP gRPC endpoint (e.g. ``http://localhost:4317``). When set, m
lmcache/v1/mp_observability/otel_init.py:44
↓ 4 callers
Method
initialize
Create + clear the slab and register it with cuFile. Args: config: GDS tier config. ``size_in_bytes`` sizes the preallocated
lmcache/v1/gpu_connector/gds_context.py:99
↓ 4 callers
Method
ipc_handle
(self)
tests/v1/test_vllm_mp_adapter.py:32
↓ 4 callers
Method
ipc_handle
(self)
lmcache/integration/vllm/vllm_multi_process_adapter.py:131
↓ 4 callers
Function
is_cuda_worker
Check if the current role is worker and a GPU accelerator is available. Args: metadata: The LMCache engine metadata. Returns:
lmcache/v1/storage_backend/__init__.py:27
↓ 4 callers
Method
is_gpu_ext_disabled
Return True when GPU extensions are disabled. Controlled by the ``NO_GPU_EXT`` environment variable. When ``True``, GPU-specific exte
setup_extensions/build_profiles/__init__.py:86
↓ 4 callers
Method
is_healthy
Check if the system is currently healthy. Returns: bool: True if healthy, False otherwise
lmcache/v1/health_monitor/base.py:261
↓ 4 callers
Method
is_worker_syncing
Check if a worker is currently in sync state. When a worker is syncing, incremental events should be discarded.
lmcache/v1/cache_controller/controllers/full_sync_tracker.py:107
↓ 4 callers
Function
keep_registered
Register, heartbeat on a timer, and deregister on cancellation. Run as an asyncio task on the MP server's event loop and cancelled on shutdow
lmcache/v1/mp_coordinator/registrar.py:75
↓ 4 callers
Function
l1_exposes_single_memory_region
Whether L1 is a single memory region a transfer channel can register. Args: config: Storage manager configuration to inspect. Return
lmcache/v1/distributed/config.py:316
↓ 4 callers
Function
loadInstances
()
lmcache/v1/cache_controller/frontend/static/js/controller_app.js:265
↓ 4 callers
Function
loadProxies
()
lmcache/lmcache_frontend/static/js/app.js:111
↓ 4 callers
Method
load_many_into
Load raw-block payloads into caller-provided memory objects. Args: encoded_keys: Ordered encoded raw-block keys to load.
lmcache/v1/storage_backend/raw_block/core.py:729
↓ 4 callers
Method
log_config
(self)
tests/cli/commands/bench/engine_bench/workloads/test_base_workload.py:27
↓ 4 callers
Function
lookup_and_wait
(adapter: DaxL2Adapter, keys: list[ObjectKey])
tests/v1/distributed/test_dax_l2_adapter.py:182
↓ 4 callers
Method
lookup_cache
Lookup the cache for the given lookup ID. Args: lookup_id: The lookup ID to lookup Returns: -1 mean
lmcache/v1/lookup_client/abstract_client.py:13
↓ 4 callers
Function
main
()
lmcache/v1/basic_check.py:75
↓ 4 callers
Function
make_raw_block_core_config
Build a small POSIX raw block core config for temp-file tests. Args: path: Backing file path for the raw block device. capacity_b
tests/v1/storage_backend/raw_block_test_utils.py:55
↓ 4 callers
Method
match_sub_sequence
Find stored chunks whose fingerprints appear anywhere in token_ids. Uses a sliding-window rolling hash so matches need not be aligned to
lmcache/v1/multiprocess/modules/blend.py:171
↓ 4 callers
Method
maybe_from_env
Build a client from ``LMCACHE_COORDINATOR_*`` env vars if configured. Reads ``LMCACHE_COORDINATOR_URL`` (required to enable) and ``LM
lmcache/v1/mp_coordinator/blend_client.py:208
↓ 4 callers
Method
maybe_submit_lookup_request
Submit a new lookup request to LMCache if there is no ongoing request. Sends a LOOKUP request to the server and blocks until a prefe
lmcache/integration/vllm/vllm_multi_process_adapter.py:700
↓ 4 callers
Function
md5
(data: bytes)
tests/v1/shm_allocator/verify_shmfile_tcp.py:139
↓ 4 callers
Method
memcheck
Perform memory check for L1 cache.
lmcache/v1/distributed/l1_manager.py:869
↓ 4 callers
Function
memory_obj_bytes
Copy a tensor memory object's byte contents into bytes. Args: obj: Tensor memory object to read. Returns: Byte copy of the o
tests/v1/storage_backend/raw_block_test_utils.py:148
↓ 4 callers
Function
merge_attention_outputs
Merge two attention outputs using LSE-based blending. Used to combine sparse attention (cached blocks) with recomputed attention (missing blo
lmcache/v1/compute/attention/triton_kernels/block_sparse_attention.py:516
↓ 4 callers
Function
normalize_raw_block_io_engine
Normalize raw-block I/O engine config with legacy compatibility. Args: io_engine: Explicit engine string. Valid values are ``"posix"``,
lmcache/v1/storage_backend/raw_block/core.py:61
↓ 4 callers
Function
num_engine_groups
Return the number of engine groups (block-id lists per transfer request). Engine group ids are assumed dense and consecutive from 0. Args:
lmcache/v1/multiprocess/group_view.py:57
↓ 4 callers
Function
objectExists
objectExists returns true if the object at key still exists in the API. It treats NotFound as the only "absent" signal; any other error is surfaced so
operator/test/utils/lmc.go:201
↓ 4 callers
Method
on_keys_created
Notify the eviction policy that new keys have been created. Args: keys (list[ObjectKey]): The keys that have been create
lmcache/v1/distributed/eviction.py:56
↓ 4 callers
Method
on_keys_touched
Notify the eviction policy that keys have been accessed. Args: keys (list[ObjectKey]): The keys that have been accessed.
lmcache/v1/distributed/eviction.py:66
↓ 4 callers
Method
on_keys_touched
Notify the eviction policy that keys have been accessed. Touched keys are moved to the most recently used position. Args:
lmcache/v1/distributed/eviction_policy/lru.py:94
↓ 4 callers
Method
on_l1_keys_deleted_by_manager
(self, keys: list[ObjectKey])
lmcache/v1/distributed/eviction.py:160
↓ 4 callers
Method
open
Register an open span and its OTel context. Args: session_id: The request session identifier. span_name: Logical name
lmcache/v1/mp_observability/subscribers/tracing/span_registry.py:50
↓ 4 callers
Function
optimal_lookup_aligned
Chain walk + block-aligned probe + chain-walk trim + fused partition. Implements Opts 1–3 from the design discussion: * Probe only at ``block
tests/v1/multiprocess/test_optimized_lookup_v3.py:254
↓ 4 callers
Function
ownedBy
ownedBy reports whether owners contains a controller reference to the CacheBlendEngine test fixture (cbeResourceName).
operator/internal/controller/cacheblendengine_controller_test.go:55
↓ 4 callers
Method
packlength
()
lmcache/v1/protocol.py:266
← previous
next →
1,501–1,600 of 13,371, ranked by callers