MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 3 callersMethod_checkpoint_once
Write a metadata checkpoint when dirty and sufficiently idle.
lmcache/v1/storage_backend/raw_block/core.py:1669
↓ 3 callersFunction_close_file_descs
Best-effort close of the FDs in descs.
lmcache/v1/storage_backend/nixl_storage_backend.py:390
↓ 3 callersFunction_config
( memory_config: L1MemoryManagerConfig, gds_l1_config: GdsL1Config | None = None, )
tests/v1/distributed/test_l1_single_region.py:18
↓ 3 callersFunction_coord_engine
A BlendV3Module mock with the coordinator-leg methods bound.
tests/v1/multiprocess/test_blend_v3_load_store_opts.py:350
↓ 3 callersMethod_correlation_key
Build the ``(session_id, device)`` correlation key. Returns ``None`` if either field is missing — such events cannot be paired safely
lmcache/v1/mp_observability/subscribers/metrics/l0_l1_throughput.py:107
↓ 3 callersFunction_create_connector
( device: torch.device, hidden_dim: int, num_layers: int, *, use_gpu: bool, chunk_size
tests/benchmarks/test_xpu_layerwise_connector_benchmark.py:70
↓ 3 callersFunction_create_connector
( device: torch.device, hidden_dim: int, num_layers: int, *, use_gpu: bool, chunk_size
tests/benchmarks/test_xpu_connector_benchmark.py:73
↓ 3 callersMethod_create_key
Convert token IDs to an IPC cache engine key. Args: token_ids: The token IDs. start: Start token index. e
lmcache/integration/vllm/vllm_multi_process_adapter.py:1006
↓ 3 callersFunction_create_local_cpu_backend
(memory_allocator)
tests/v1/shm_allocator/test_shm_allocator.py:121
↓ 3 callersFunction_create_shm_segment
Create a POSIX SHM segment via the project facade. Returns the owner mmap address so the test can release the segment with ``shm_munmap`` + `
tests/v1/multiprocess/test_engine_driven_transfer.py:1366
↓ 3 callersFunction_create_shmfs_config
Build config with shmfs.* extra_config entries.
tests/v1/shm_allocator/test_shm_allocator.py:101
↓ 3 callersFunction_detect_with_stub
Run ``_detect_device`` with ``torch`` swapped for the stub.
tests/v1/test_musa_support.py:113
↓ 3 callersFunction_device_from_type
(device_type)
tests/benchmarks/test_xpu_layerwise_connector_benchmark.py:49
↓ 3 callersFunction_device_from_type
(device_type: str)
tests/benchmarks/test_xpu_connector_benchmark.py:37
↓ 3 callersMethod_device_status_locked
(self, index: int, entry: DaxDeviceEntry)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1248
↓ 3 callersFunction_directory
Return the shared blend directory from app state. Args: request: The incoming request. Returns: The shared :class:`GlobalBle
lmcache/v1/mp_coordinator/http_apis/blend_directory_api.py:35
↓ 3 callersFunction_dtype_to_int
Local re-export of the protocol DTYPE_TO_INT to avoid a circular import at module-load time.
lmcache/v1/kv_codec/encoded_kv.py:214
↓ 3 callersFunction_emit_final_metrics
Emit final benchmark summary using the CLI metrics system.
lmcache/cli/commands/bench/engine_bench/command.py:448
↓ 3 callersMethod_ensure_hotplug_enabled
(self)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1555
↓ 3 callersMethod_ensure_metadata
(self)
lmcache/integration/vllm/vllm_service_factory.py:235
↓ 3 callersMethod_ensure_open_locked
(self)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1007
↓ 3 callersMethod_execute_operation
Execute a single operation Returns: Tuple of (message_count, request_count, latency, error)
lmcache/tools/controller_benchmark/benchmark.py:449
↓ 3 callersMethod_executor_submit_lease_acquisition
Submit lease acquisition to priority executor.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:407
↓ 3 callersMethod_exists_sync
(self, key_str: str)
lmcache/v1/storage_backend/connector/eic_connector.py:260
↓ 3 callersFunction_expected_kernel_group_bytes
Byte size of one kernel group's per-chunk buffer.
tests/v1/platform/test_gpu_cache_context.py:149
↓ 3 callersFunction_fake_stream
A stand-in for ``torch_dev.current_stream()`` (no CUDA needed).
tests/v1/gpu_connector/test_gds_context.py:35
↓ 3 callersMethod_fetch_object_sizes_sync
Fetch exact object sizes for ``key_strings`` in one metadata request.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:638
↓ 3 callersMethod_format_object_key_b128
Generate a 128-bit object key for the DOCA_MEMOS backend. Returns a 32-char hex string (sha256 truncated to 128 bits). Hex encoding
lmcache/v1/storage_backend/nixl_storage_backend.py:1500
↓ 3 callersMethod_format_object_key_url_safe
Generate a URL-safe object key for non-DOCA_MEMOS backends. Replaces slashes and @ signs with underscores, then percent-encodes
lmcache/v1/storage_backend/nixl_storage_backend.py:1510
↓ 3 callersFunction_format_timestamp
Format a unix timestamp as a compact datetime string. Example: 20260401_143025
lmcache/v1/mp_observability/subscribers/logging/lookup_hash.py:34
↓ 3 callersFunction_gbps
Return throughput in GB/s (decimal GB) for ``num_bytes`` over ``seconds``.
lmcache/tools/transfer_channel_benchmark/benchmark.py:68
↓ 3 callersMethod_generate_query
Generate a user query of approximately ``user_input_length`` tokens.
lmcache/cli/commands/bench/engine_bench/workloads/multi_round_chat.py:234
↓ 3 callersFunction_generate_random_body
Build a body of ``num_tokens`` BPE tokens of unique random content. Samples random token-IDs in the safe range and decodes them. The decoded
lmcache/cli/commands/bench/engine_bench/workloads/prefix_suffix_tuner.py:344
↓ 3 callersMethod_get_chunk_block_ids
( self, block_ids: List[int], start: int )
lmcache/v1/gpu_connector/gpu_connectors.py:2085
↓ 3 callersMethod_get_device_uuid
Get the UUID of a device given its index.
lmcache/v1/multiprocess/custom_types.py:47
↓ 3 callersFunction_get_engine
Extract the LMCache engine from the request state. Returns: A tuple of (engine, error_response). If engine is None, error_re
lmcache/v1/internal_api_server/vllm/hot_cache_api.py:13
↓ 3 callersMethod_get_free_chunks
Return number of free chunks in the PD buffer allocator. Reads ``free_blocks`` from the appropriate underlying allocator (CPU or GPU
lmcache/v1/storage_backend/pd_backend_async.py:1543
↓ 3 callersMethod_get_mapped_device_locked
(self, key: ObjectKey)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1193
↓ 3 callersFunction_get_metadata
()
tests/v1/shm_allocator/test_shm_allocator.py:87
↓ 3 callersMethod_get_next_task_id
(self)
lmcache/v1/distributed/l2_adapters/nixl_store_dynamic_l2_adapter.py:575
↓ 3 callersMethod_get_next_task_id
Get the next task ID and increment the counter.
lmcache/v1/distributed/l2_adapters/nixl_store_l2_adapter.py:684
↓ 3 callersMethod_get_next_task_id
(self)
lmcache/v1/distributed/l2_adapters/fs_l2_adapter.py:472
↓ 3 callersMethod_get_next_task_id
Get the next task ID and increment the counter.
lmcache/v1/distributed/l2_adapters/mock_l2_adapter.py:340
↓ 3 callersMethod_get_next_task_id_locked
(self)
lmcache/v1/distributed/l2_adapters/hfbucket_l2_adapter.py:519
↓ 3 callersMethod_get_next_task_id_locked
(self)
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1011
↓ 3 callersMethod_get_next_task_id_locked
(self)
lmcache/v1/distributed/l2_adapters/raw_block_l2_adapter.py:612
↓ 3 callersFunction_get_object_path
Safely read the ``path`` attribute from a Hugging Face path object.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:918
↓ 3 callersMethod_get_object_size_async
(self, key_str: str)
lmcache/v1/storage_backend/connector/s3_connector.py:229
↓ 3 callersMethod_get_positive_int
Extract a positive integer from config with validation. Args: key: Configuration key default: Default value i
lmcache/v1/storage_backend/connector/hf3fs_adapter.py:108
↓ 3 callersMethod_get_positive_int_extra
( extra_config: dict[str, Any], key: str, default: int, )
lmcache/v1/storage_backend/plugins/dax_backend.py:613
↓ 3 callersMethod_get_request_tracker
(self, request_id: str)
lmcache/integration/vllm/lmcache_mp_connector_0180.py:1040
↓ 3 callersFunction_get_role
Get the role from the manager's engine metadata. Returns "scheduler", "worker", or "unknown".
lmcache/v1/internal_api_server/vllm/lookup_api.py:22
↓ 3 callersFunction_get_server_url
Resolve the server URL: connector-config field > env var > default.
lmcache/integration/tensorrt_llm/tensorrt_mp_adapter.py:50
↓ 3 callersMethod_get_shape_for_kernel_group
Returns the shape of the temp GPU buffer for the given kernel group index Args: num_tokens: Number of tokens
lmcache/v1/platform/cuda/cache_context.py:262
↓ 3 callersFunction_get_statistics_client
( request: Request, )
lmcache/v1/internal_api_server/vllm/chunk_statistics_api.py:44
↓ 3 callersFunction_get_storage_manager
Get the StorageManager from the request's app state. Returns: Tuple of (storage_manager, error_response). If storage_manager
lmcache/v1/internal_api_server/vllm/bypass_api.py:10
↓ 3 callersFunction_get_storage_manager
(request: Request)
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:95
↓ 3 callersMethod_handle_alloc_request
Handle a single allocation or cancellation request from a sender. Dispatches AllocRequest to _async_allocate_and_put. CancelNotif rel
lmcache/v1/storage_backend/pd_backend_async.py:1210
↓ 3 callersMethod_hash_tokens
( self, tokens: Union[torch.Tensor, List[int]], prefix_hash: Optional[int] = None,
lmcache/v1/token_database.py:268
↓ 3 callersMethod_hashes
(self, item: Union[str, int])
lmcache/v1/utils/bloom_filter.py:37
↓ 3 callersMethod_http_request
Execute HTTP request with optional semaphore gate.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:218
↓ 3 callersFunction_info
(msg: str)
lmcache/cli/commands/query/_request.py:34
↓ 3 callersMethod_init_health_monitor
Initialize the health monitor via the service factory. Called during post_init after all components are initialized.
lmcache/v1/manager.py:215
↓ 3 callersFunction_install
Replace the live backend table with *backends*.
tests/v1/platform/test_cache_context_dispatch.py:141
↓ 3 callersFunction_instance
(instance_id: str, ip: str = "10.0.0.1", port: int = 8000)
tests/v1/mp_coordinator/test_eviction_manager.py:234
↓ 3 callersFunction_instance
(instance_id: str, heartbeat: float)
tests/v1/mp_coordinator/test_health.py:12
↓ 3 callersFunction_int_to_dtype
(idx: int)
lmcache/v1/kv_codec/encoded_kv.py:227
↓ 3 callersFunction_is_not_found_error
Return whether an exception represents a missing bucket/object.
lmcache/v1/distributed/l2_adapters/hfbucket_l2_adapter.py:131
↓ 3 callersMethod_key_to_string
Convert CacheEngineKey to URL-safe string format.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:294
↓ 3 callersFunction_layerwise_store_vllm_contract
Mimic vLLM layerwise store contract: - create one store_layer generator per request (first one sync=True, rest sync=False) -
tests/benchmarks/test_xpu_layerwise_connector_benchmark.py:91
↓ 3 callersMethod_log_kvcache_for_check
Helper method to log KVCache Check information. This method centralizes the KVCache Check logging logic that was duplicated
lmcache/v1/cache_engine.py:1947
↓ 3 callersMethod_lookup
(self, adapter: HFBucketL2Adapter, key: ObjectKey)
tests/v1/distributed/test_hfbucket_l2_adapter.py:356
↓ 3 callersMethod_lookup
(self, adapter, key)
tests/v1/distributed/test_s3_l2_adapter.py:534
↓ 3 callersMethod_make
(self)
tests/v1/multiprocess/test_optimized_lookup_v3.py:692
↓ 3 callersMethod_make_cache_key
(self, mm_hash: str)
lmcache/v1/ec_engine.py:115
↓ 3 callersFunction_make_config
Create a minimal mock config for testing.
tests/test_observability.py:303
↓ 3 callersFunction_make_ctx
(wait_result=True, found=None)
tests/v1/multiprocess/test_lookup_wait_prefetch.py:19
↓ 3 callersFunction_make_encode_inputs
(xops, nlayers, ntokens, nchannels, max_bins)
tests/benchmarks/test_xpu_kernels_microbench.py:79
↓ 3 callersFunction_make_engine_with_mocked_matcher
Construct a real BlendV3Module with the matcher mocked so we can observe `on_new_token_hashes` calls without setting up storage.
tests/v1/multiprocess/test_blend_v3_load_store_opts.py:24
↓ 3 callersMethod_make_full_mask_csr
CSR for a fully dense mask (all blocks attended).
tests/v1/compute/attention/test_triton_kernels.py:141
↓ 3 callersFunction_make_key
Unique CacheEngineKey for tests.
tests/v1/storage_backend/test_nixl_shared_pool.py:260
↓ 3 callersMethod_make_key
Create a valid CacheEngineKey for testing.
tests/v1/storage_backend/test_pd_backend_bidirectional.py:385
↓ 3 callersFunction_make_module
(ctx)
tests/v1/multiprocess/test_lookup_wait_prefetch.py:30
↓ 3 callersFunction_make_tensor_obj
Create a TensorMemoryObj compatible with both old and new API.
examples/lmc_external_l2_adapter/tests/test_plugin.py:85
↓ 3 callersMethod_meta_container_offsets
Return byte offsets for mirrored metadata checkpoint containers.
lmcache/v1/storage_backend/raw_block/core.py:1499
↓ 3 callersMethod_metadata_to_labels
(metadata: LMCacheMetadata)
lmcache/observability.py:1905
↓ 3 callersFunction_native_mooncake_available
Check if the C++ Mooncake extension can be imported.
tests/v1/distributed/test_mooncake_store_l2_adapter.py:51
↓ 3 callersFunction_next_url
()
tests/v1/distributed/l2_adapters/test_p2p_l2_adapter_integration.py:69
↓ 3 callersFunction_normalize_ec_config_key
Normalize one EC-prefixed config key to a LMCache config key.
lmcache/v1/config.py:1085
↓ 3 callersMethod_notify_staging_freed
Wake threads blocked in allocate() so they can retry after RDMA frees memory. Called from _async_transfer_task finally block (on sender event
lmcache/v1/storage_backend/pd_backend_async.py:985
↓ 3 callersMethod_on_request_failure
Create a default return message when worker -> controller request encounters an error (e.g., timeout, ZMQ error).
lmcache/v1/cache_controller/worker.py:353
↓ 3 callersFunction_package_resource_path
Return absolute filesystem path for a file shipped inside the package. Replacement for the deprecated ``pkg_resources.resource_filename``; wo
lmcache/lmcache_frontend/app.py:23
↓ 3 callersFunction_parse_quoted_string
Parse a string that may be surrounded by quotes and handle escape characters. Args: value: The input string that may be quoted Retur
lmcache/v1/config_base.py:137
↓ 3 callersMethod_pin_if_needed
(self, encoded_key: str)
lmcache/v1/storage_backend/plugins/rust_raw_block_backend.py:669
↓ 3 callersFunction_prompt_for_item
Prompt the user for a single config item based on its type.
lmcache/cli/commands/bench/engine_bench/interactive/__init__.py:39
↓ 3 callersFunction_query_checksum
Query KV cache checksums via the HTTP API. Uses the MP-native ``block_ids`` + ``block_size`` addressing scheme so the query matches the same
lmcache/cli/commands/bench/server_bench/helpers.py:826
↓ 3 callersMethod_raise_if_closed_locked
(self)
lmcache/v1/distributed/l2_adapters/raw_block_l2_adapter.py:608
↓ 3 callersFunction_read_document_file
(path: str, *, name: Optional[str] = None)
lmcache/cli/commands/query/_prompt.py:116
↓ 3 callersFunction_read_histograms
()
tests/v1/mp_observability/subscribers/metrics/test_l0_l1_throughput.py:76
↓ 3 callersFunction_read_histograms
()
tests/v1/mp_observability/subscribers/metrics/test_l2_throughput.py:124
← previousnext →1,701–1,800 of 13,371, ranked by callers