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
↓ 5 callers
Method
_make_agent
()
tests/v1/test_nixl_batched_contains.py:78
↓ 5 callers
Function
_make_byte_obj
(size: int)
tests/v1/storage_backend/test_rust_raw_block_backend.py:112
↓ 5 callers
Function
_make_cli_args
Build a Namespace mimicking bench engine CLI output.
tests/cli/commands/bench/engine_bench/interactive/test_state.py:32
↓ 5 callers
Function
_make_connector
(requests)
tests/v1/test_vllm_layerwise_wait_for_save.py:64
↓ 5 callers
Method
_make_key
Create a valid CacheEngineKey for testing.
tests/v1/storage_backend/test_pd_backend_bidirectional.py:174
↓ 5 callers
Function
_make_layout_desc
()
tests/v1/multiprocess/test_p2p_controller.py:43
↓ 5 callers
Function
_make_mock_metrics
Build a (mock_metrics_module, mock_meter) pair for patching OTel.
tests/v1/mp_observability/subscribers/metrics/test_event_bus_self_metrics.py:23
↓ 5 callers
Function
_make_mock_result
(request_id: str = "req_0")
tests/cli/commands/bench/engine_bench/workloads/test_prefix_suffix_tuner.py:211
↓ 5 callers
Method
_make_mocked_backend
Return a GdsBackend with tmpfs mocked so GDS is auto-disabled.
tests/v1/storage_backend/test_gds_backend.py:512
↓ 5 callers
Function
_make_module_with_job
Create a LookupModule with a mock context and a single prefetch job. Returns: (module, request_id)
tests/v1/multiprocess/test_query_lookup_hits.py:109
↓ 5 callers
Function
_make_run_script_app
(state_kwargs: dict)
tests/v1/multiprocess/http_apis/test_common_api.py:74
↓ 5 callers
Function
_make_shape_desc
Build a ``PageBufferShapeDesc`` with the given geometry.
tests/v1/gpu_connector/test_concrete_shape.py:16
↓ 5 callers
Function
_make_sm
(adapters: list[_StubAdapter], names: list[str])
tests/v1/distributed/test_storage_manager_l2_keys.py:37
↓ 5 callers
Function
_make_unique_slot_mapping
( *, total_slots: int, num_tokens: int, device: torch.device )
tests/v1/test_musa_connector.py:35
↓ 5 callers
Function
_make_worker_adapter
Construct a worker adapter with the standard test arguments; the network boundary must already be patched (see ``fake_adapter``). ``extra_conf
tests/v1/test_vllm_mp_adapter.py:95
↓ 5 callers
Function
_matcher_request
request_fn that drives a real matcher, mirroring the coordinator router.
tests/v1/mp_coordinator/test_blend_client.py:31
↓ 5 callers
Function
_object_key_to_filename
Build a reversible, filesystem-safe filename. Unsalted:: <safe_model>@0x<kv_rank_hex>@<object_group_id_hex>@<chunk_hash_hex>.data S
lmcache/v1/distributed/l2_adapters/fs_l2_adapter.py:98
↓ 5 callers
Function
_parse_mp_storage_args
(args: list[str])
tests/v1/distributed/test_devdax_l1_allocator.py:65
↓ 5 callers
Method
_patch_base
Make *adapter_cls* pass issubclass check.
tests/v1/distributed/test_l2_adapter_factory.py:339
↓ 5 callers
Function
_patch_musa_connector_attrs
Patch connector layout discovery so transfer logic can run on CPU.
tests/v1/test_musa_connector.py:70
↓ 5 callers
Function
_percentiles
(values: list[float], pcts: list[int])
lmcache/tools/cache_simulator/simulator.py:324
↓ 5 callers
Method
_pool_size_gb_to_bytes
Convert pool size in GB to bytes.
lmcache/v1/storage_backend/maru_backend.py:97
↓ 5 callers
Function
_raw_blocks_first_caches
Per-layer blocks-first tensors as registered: [NB, NH, BS, 2 * HS].
tests/v1/gpu_connector/test_blocks_first_fused_kv_format.py:43
↓ 5 callers
Method
_read_buffers
Read one or more buffers through the configured Rust I/O path. Args: offsets: Device offsets for each read. buffers:
lmcache/v1/storage_backend/raw_block/core.py:1299
↓ 5 callers
Method
_record_connection_outcome
Update the circuit breaker under the lock.
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:1014
↓ 5 callers
Function
_resolve_config_aliases
Resolve configuration aliases and handle deprecated configurations.
lmcache/v1/config_base.py:184
↓ 5 callers
Function
_run
()
tests/v1/storage_backend/test_pd_backend_async.py:514
↓ 5 callers
Function
_skip_if_no_musa
()
tests/v1/test_musa_connector.py:30
↓ 5 callers
Function
_slashed
Inverse of :func:`_strip_leading_slash` for shm_open calls.
lmcache/v1/multiprocess/posix_shm.py:58
↓ 5 callers
Method
_snapshot_adapters
Snapshot the active adapters under the lock, in ascending adapter-id order. Iterate this instead of the live dicts so a concurrent add
lmcache/v1/distributed/storage_manager.py:975
↓ 5 callers
Method
_status_response
( self, cache_size_per_token: int = 163840, world_size: int = 1, model_name: s
tests/cli/commands/bench/engine_bench/test_config.py:237
↓ 5 callers
Function
_tensor_to_bytes_fast
Convert a contiguous CPU tensor to bytes via a single C-level memcpy. Reinterpreting the tensor as uint8 and calling numpy's ``.tobytes()`` e
lmcache/v1/kv_codec/asym_k16_v8.py:43
↓ 5 callers
Function
_valid_block_range_indices
Return valid [start, end) range over flat block IDs and object token offset.
lmcache/python_ops_fallback.py:1263
↓ 5 callers
Method
acquire
(self, n: int = 1)
lmcache/v1/storage_backend/storage_manager.py:131
↓ 5 callers
Method
acquire_read
Acquire a read lock with optional timeout.
lmcache/v1/cache_controller/locks.py:35
↓ 5 callers
Method
add_affinity_thread_pool
Assign an AffinityThreadPool to specific request types. Use this for GPU-bound blocking handlers (e.g. STORE, RETRIEVE). Requests fro
lmcache/v1/multiprocess/mq.py:818
↓ 5 callers
Function
add_mp_server_args
Add MP server configuration arguments to an existing parser. Args: parser: The argument parser to add arguments to. Returns:
lmcache/v1/multiprocess/config.py:218
↓ 5 callers
Function
applyArg
applyArg appends-or-replaces a "--flag value" pair in args (design §9.1). It recognizes both the two-token form ["--flag", "value"] and the single-tok
operator/internal/webhook/cacheblend_inject_builders.go:239
↓ 5 callers
Function
apply_mm_hashes_to_token_ids
Overwrite token_ids in-place for multimodal placeholders using efficient slice assignments.
lmcache/integration/vllm/utils.py:171
↓ 5 callers
Function
argsHasFlag
argsHasFlag reports whether args already carries the given flag in either the two-token "--flag value" form or the single-token "--flag=value" form.
operator/internal/webhook/pod_injector.go:338
↓ 5 callers
Function
auto_detect_model
Fetch the first model ID from the engine's ``/v1/models`` endpoint. Args: engine_url: Base URL of the inference engine (e.g.,
lmcache/cli/commands/bench/engine_bench/config.py:58
↓ 5 callers
Method
batched_get_blocking
Synchronously load the leading raw-block hit prefix. Args: keys: Ordered legacy cache keys to load. Returns:
lmcache/v1/storage_backend/plugins/rust_raw_block_backend.py:585
↓ 5 callers
Method
batched_op
Create BatchedKVOperationMsg.
tests/v1/cache_controller/conftest.py:192
↓ 5 callers
Method
batched_p2p_lookup
Perform batched P2P lookup for multiple keys. :param BatchedP2PLookupMsg msg: The batched P2P lookup message containing keys.
lmcache/v1/cache_controller/controllers/kv_controller.py:403
↓ 5 callers
Method
batched_read
Read a batch of data from a remote peer through the nixl channel. This is the reverse of batched_write: the local side initiates a R
lmcache/v1/transfer_channel/nixl_channel.py:462
↓ 5 callers
Method
batched_write
Write a batch of data through the nixl channel. :param objects: A list of bytes or MemoryObj to be written. :param transfer_
lmcache/v1/transfer_channel/nixl_channel.py:419
↓ 5 callers
Function
bench_case
Print Python vs native timings for one (size, fill) configuration.
benchmarks/microbenchmark/bitmap_ops_benchmark.py:43
↓ 5 callers
Function
block_sparse_attention
Compute block-sparse attention. Args: query: [M, H, D] query tensor key: [N, H, D] key tensor value: [N, H, D] value tens
lmcache/v1/compute/attention/triton_kernels/block_sparse_attention.py:390
↓ 5 callers
Function
build_format
(color)
lmcache/logging.py:8
↓ 5 callers
Function
build_format
(color)
benchmarks/multi_round_qa/utils.py:9
↓ 5 callers
Function
build_format
(color)
benchmarks/rag/utils.py:17
↓ 5 callers
Method
build_messages
Construct OpenAI-format messages for a request.
lmcache/cli/commands/bench/engine_bench/workloads/multi_round_chat.py:121
↓ 5 callers
Method
bytes_transferred
Return the number of bytes actually written, or 0 on failure.
lmcache/v1/distributed/internal_api.py:198
↓ 5 callers
Method
calculate_num_blocks
Calculate the number of blocks for a given number of tokens in a specified kernel group. Args: kernel_group_idx: 0-based
lmcache/v1/kv_layer_groups.py:533
↓ 5 callers
Method
check
(self)
tests/v1/test_health_monitor.py:38
↓ 5 callers
Function
check_gpu_and_cpu_equal
( gpu_tensor: torch.Tensor, cpu_tensor: torch.Tensor, )
tests/v1/test_mem_kernels.py:745
↓ 5 callers
Method
close
Sync the stream, deregister cuFile state, and close the slab handle.
lmcache/v1/gpu_connector/gds_context.py:211
↓ 5 callers
Method
close
(self)
lmcache/v1/distributed/l2_adapters/native_connector_l2_adapter.py:371
↓ 5 callers
Method
close
(self)
lmcache/v1/distributed/l2_adapters/mock_l2_adapter.py:248
↓ 5 callers
Method
close
(self)
lmcache/v1/storage_backend/connector/hf3fs_connector.py:566
↓ 5 callers
Method
close
Close the underlying file. Idempotent.
lmcache/v1/mp_observability/trace/reader.py:102
↓ 5 callers
Method
contains
Report whether an instance is currently registered. Args: instance_id: Identifier to check. Returns: ``True`
lmcache/v1/mp_coordinator/registry.py:110
↓ 5 callers
Function
countLogLines
countLogLines is the implementation behind countLMCacheStored / countLMCacheRetrieved. We shell out to kubectl rather than building a typed Pods/log c
operator/test/e2e/vllm_integration_smoke_test.go:259
↓ 5 callers
Function
create_serde_processor
Build a SerdeProcessor from a SerdeConfig. Args: config: Serde configuration. ``config.type`` must name a registered serde; `
lmcache/v1/distributed/serde/factory.py:52
↓ 5 callers
Function
create_test_config
Create a test configuration for NixlStorageBackend
tests/disagg/test_nixl_storage_backend.py:64
↓ 5 callers
Function
create_test_config
Create test configuration for P2P backend
tests/v1/storage_backend/test_p2p_backend_with_controller.py:173
↓ 5 callers
Function
create_test_key
Create a test CacheEngineKey.
lmcache/v1/check/utils.py:85
↓ 5 callers
Function
create_test_metadata
Create test metadata for NixlStorageBackend
tests/disagg/test_nixl_storage_backend.py:81
↓ 5 callers
Function
create_test_metadata
(kv_shape=(2, 2, 256, 8, 128))
tests/v1/storage_backend/test_bigtable_integration.py:24
↓ 5 callers
Function
create_test_metadata
Create a minimal test metadata instance.
tests/v1/storage_backend/test_local_cpu_backend_nixl_paged.py:24
↓ 5 callers
Function
create_test_metadata
Create test metadata
tests/v1/storage_backend/test_p2p_backend_with_controller.py:196
↓ 5 callers
Method
cudaHostUnregister
(self, ptr: int)
tests/v1/distributed/test_devdax_l1_allocator.py:97
↓ 5 callers
Method
cudart
(self)
tests/v1/distributed/test_devdax_l1_allocator.py:90
↓ 5 callers
Method
deregister
Remove an mp server entry and return it. Args: instance_id: Identifier of the instance to remove. Returns: T
lmcache/v1/mp_coordinator/registry.py:85
↓ 5 callers
Method
device_count
Return the number of available devices. Returns: 1 always for the stub.
lmcache/v1/platform/cpu/stub_cpu_device.py:290
↓ 5 callers
Function
escape_salt
Translate the empty-string salt to the URL sentinel.
lmcache/cli/commands/quota/_helpers.py:28
↓ 5 callers
Method
estimate_serialized_size
( self, layout_descs: LayoutDescGroup, )
lmcache/v1/distributed/serde/asym_k16_v8.py:394
↓ 5 callers
Method
execute
(self, args: argparse.Namespace)
lmcache/cli/commands/describe.py:327
↓ 5 callers
Method
execute_evictions
Compute the plan and fire-and-forget a ``DELETE /l2`` to one random registered MP server. Returns the scheduled plan as soon as the b
lmcache/v1/mp_coordinator/l2/eviction_manager.py:122
↓ 5 callers
Function
fmt_health
Format a boolean health flag as ``'OK'`` / ``'UNHEALTHY'``.
lmcache/cli/commands/describe.py:72
↓ 5 callers
Function
fold
Fold per-(group, chunk, rank) presence into servable prefix lengths. For each object group, computes which prefix lengths it can serve under its
lmcache/v1/distributed/bitmap_ops/fold.py:67
↓ 5 callers
Method
free
(self, memory_obj: MemoryObj, allocator_type: Optional[str] = None)
lmcache/v1/memory_management.py:2341
↓ 5 callers
Method
get
Blocking function to get the memory object from the storages.
lmcache/v1/storage_backend/storage_manager.py:434
↓ 5 callers
Function
getServerPort
(spec *lmcachev1alpha1.LMCacheEngineSpec)
operator/internal/resources/compute.go:189
↓ 5 callers
Method
get_all
Return a snapshot copy of per-salt byte usage.
lmcache/v1/mp_coordinator/l2/usage_manager.py:100
↓ 5 callers
Method
get_all_worker_infos_cached
Get WorkerInfo for all workers across all instances. Args: timeout_seconds: Cache timeout in seconds. - If None (
lmcache/v1/cache_controller/utils.py:485
↓ 5 callers
Function
get_available_port
Get an available port dynamically by binding to port 0. Args: host: The host address to bind to. Default is "127.0.0.1". Return
tests/v1/utils.py:237
↓ 5 callers
Function
get_available_ports
Get multiple available ports dynamically. Args: count: Number of ports to get. host: The host address to bind to. Default is
tests/v1/utils.py:254
↓ 5 callers
Function
get_block_data
Extract all layer data for a given block.
tests/v1/test_mp_mem_kernels.py:168
↓ 5 callers
Method
get_event_status
Get the status of the event with the given type and id.
lmcache/v1/event_manager.py:95
↓ 5 callers
Method
get_file_path_for_key
Return the full file path for a given ObjectKey.
lmcache/v1/distributed/l2_adapters/nixl_store_dynamic_l2_adapter.py:281
↓ 5 callers
Function
get_hidden_dim_size
Return the hidden dimension for a layer (defaults to layer 0).
lmcache/v1/gpu_connector/utils.py:282
↓ 5 callers
Method
get_l1_memory_desc
Return an L1MemoryDesc describing the underlying L1 memory buffer.
lmcache/v1/distributed/l1_manager.py:814
↓ 5 callers
Method
get_load_event_fd
Get the event fd for load operation, which will be signaled on the completion of the load tasks. Returns: int: t
lmcache/v1/distributed/l2_adapters/base.py:182
↓ 5 callers
Method
get_lookup_and_lock_event_fd
(self)
examples/lmc_external_l2_adapter/src/lmc_external_l2_adapter/adapter.py:152
↓ 5 callers
Method
get_lookup_and_lock_event_fd
(self)
lmcache/v1/distributed/l2_adapters/nixl_store_dynamic_l2_adapter.py:391
↓ 5 callers
Method
get_lookup_and_lock_event_fd
Get the event fd for lookup and lock operation, which will be signaled on the completion of the lookup and lock tasks. Retur
lmcache/v1/distributed/l2_adapters/base.py:167
↓ 5 callers
Method
get_memory_allocator
(self)
lmcache/v1/storage_backend/local_cpu_backend.py:938
↓ 5 callers
Method
get_missing_batches
Get list of missing batch IDs that need to be resent. Args: instance_id: The instance ID worker_id: The work
lmcache/v1/cache_controller/controllers/full_sync_tracker.py:416
↓ 5 callers
Method
get_non_allocator_backends
Get the names of the actual storage backends. Some backends, such as LocalCPUBackend and PDBackend, in some cases, only serve
lmcache/v1/storage_backend/storage_manager.py:1191
← previous
next →
1,201–1,300 of 13,371, ranked by callers