MCPcopy Create free account

hub / github.com/LMCache/LMCache / functions

Functions13,371 in github.com/LMCache/LMCache

↓ 1 callersFunction_merge_layer_sw_sizes
Merge the per-layer sliding window sizes of one LMCache group. Args: per_layer_sw_size: Sliding window size per registered tensor index.
lmcache/integration/vllm/kv_cache_groups.py:65
↓ 1 callersFunction_metadata
()
tests/v1/test_hidden_state_store.py:28
↓ 1 callersMethod_metadata_to_key
(metadata: LMCacheMetadata)
lmcache/observability.py:1915
↓ 1 callersMethod_migrate_keys_locked
( self, source: DaxDeviceEntry, keys: list[ObjectKey], )
lmcache/v1/distributed/l2_adapters/dax_l2_adapter.py:1323
↓ 1 callersFunction_native_client_available
Check if the C++ Redis extension can be imported.
tests/v1/distributed/test_resp_l2_adapter_integration.py:47
↓ 1 callersFunction_native_extension_available
()
tests/v1/distributed/test_aerospike_l2_integration.py:51
↓ 1 callersFunction_native_transfer_dims
Return ``(num_heads, head_size)`` for native MUSA transfer.
lmcache/v1/platform/musa/native_kv_transfer.py:360
↓ 1 callersFunction_needs_interactive
Check whether interactive mode should be triggered.
lmcache/cli/commands/bench/engine_bench/command.py:346
↓ 1 callersMethod_needs_rotation
Check if the current file needs rotation.
lmcache/v1/mp_observability/subscribers/logging/lookup_hash.py:171
↓ 1 callersFunction_nixl_gpu_config
(pool_size: int = 0)
tests/v1/storage_backend/test_nixl_shared_pool.py:143
↓ 1 callersMethod_nixl_transfer_async
Memory allocation and NIXL transfer without locks. Can run async and in parallel with other transfers.
lmcache/v1/storage_backend/nixl_storage_backend.py:1124
↓ 1 callersMethod_non_overlapping_after_prefix
Matches outside the prefix coverage, leftmost-greedy overlap-deduped. Drops matches the prefix leg already covers (``cur_st < prefix_tokens``
lmcache/v1/multiprocess/modules/blend_v3.py:570
↓ 1 callersFunction_normalize_levels
Normalize the ``levels`` argument to a ``(min_depth, max_depth)`` pair. ``max_depth == 0`` is the sentinel for "no upper bound". Accepted for
lmcache/v1/utils/subclass_discovery.py:38
↓ 1 callersFunction_normalize_lmcache_objects
Normalize LMCache object inputs to chunk tensors. Accepts either a list of chunk tensors or a ``list[int]`` of raw CPU pointers. When a point
lmcache/python_ops_fallback.py:1034
↓ 1 callersFunction_normalize_operation
(operation: str)
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:157
↓ 1 callersFunction_normalize_paged_layers
Normalize paged buffer input based on GPU KV format. Accepts either tensor-form inputs (list / Tensor) or a 1-D pointer tensor (int64 / uint6
lmcache/python_ops_fallback.py:906
↓ 1 callersFunction_normalize_save_chunk_meta_config
Clone config and default ``save_chunk_meta`` to ``False`` for hfbucket. The generic ``RemoteConnector`` base defaults ``save_chunk_meta`` to ``Tr
lmcache/v1/storage_backend/connector/hfbucket_connector.py:833
↓ 1 callersFunction_object_key
(suffix: int)
tests/v1/distributed/test_aerospike_l2_integration.py:99
↓ 1 callersMethod_on_load_completed
(self, event: Event)
lmcache/v1/mp_observability/subscribers/metrics/l2.py:175
↓ 1 callersMethod_on_peer_connected
Hook for subclasses to perform additional setup when a peer connects
lmcache/v1/transfer_channel/py_socket_channel.py:199
↓ 1 callersMethod_open_and_register_slab
Create, truncate, preallocate the slab file and register it with cuFile. Args: use_direct_io: Open with ``O_DIRECT`` (required fo
lmcache/v1/gpu_connector/gds_context.py:244
↓ 1 callersMethod_open_devdax_mapping
( self, device_path: str, size: int, )
lmcache/v1/memory_management.py:2634
↓ 1 callersFunction_openai_error
(obj: dict[str, Any])
lmcache/cli/commands/query/_request.py:38
↓ 1 callersFunction_operation_payload
( backend: str, operation: str, payload: dict[str, object], )
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:308
↓ 1 callersMethod_optimal_hash_count
(m: int, n: int)
lmcache/v1/utils/bloom_filter.py:34
↓ 1 callersMethod_optimal_size
(n: int, p: float)
lmcache/v1/utils/bloom_filter.py:30
↓ 1 callersFunction_os_flags
Translate a Python-style mode string to ``os.open`` flags.
lmcache/v1/storage_backend/hipfile_shim.py:20
↓ 1 callersFunction_parse_docstring_params
Parse pybind11 docstring to extract [(name, has_default, default_value)]. pybind11 generates docstrings like: func_name(arg1: type1, arg2
tests/v1/test_c_ops_fallback_parity.py:132
↓ 1 callersFunction_parse_list_query
Pull ``prefix`` / ``max-keys`` / ``continuation-token`` out of the ListObjectsV2 query string.
tests/v1/distributed/test_s3_l2_adapter.py:145
↓ 1 callersFunction_parse_list_response_xml
Parse a ListObjectsV2 XML response into ``(entries, next_token)``. Entries this adapter can't parse (foreign objects in the bucket) are skipp
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:118
↓ 1 callersFunction_parse_mp_config
()
lmcache/lmcache_frontend/lmcache_mp_plugin/lmcache_mp_frontend_plugin.py:75
↓ 1 callersMethod_parse_per_op_workers_from_dict
Parse ``per_op_workers`` from a raw configuration dict. Returns ``None`` if the key is absent. Raises: ValueError: If th
lmcache/v1/distributed/l2_adapters/config.py:311
↓ 1 callersMethod_parse_shape
Parse shape from metadata, removing padding zeros.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:623
↓ 1 callersFunction_parse_size_string
(size: str)
lmcache/v1/multiprocess/http_apis/reconfigure_api.py:105
↓ 1 callersFunction_parse_tokens_from_params
Parse tokens from input parameters. Args: tokens_mock: Two comma-separated numbers specifying start and end of token range -
lmcache/v1/internal_api_server/vllm/cache_api.py:28
↓ 1 callersMethod_parse_url
Parse and normalize the SageMaker HyperPod URL. Args: url: Raw URL from context (e.g., "sagemaker-hyperpod://127.0.0.1:9
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_adapter.py:155
↓ 1 callersFunction_path_to_key
(path: str)
tests/v1/distributed/test_s3_l2_adapter.py:96
↓ 1 callersMethod_peer_changed
Return whether a peer's advertised connection info changed. Args: adapter: The currently tracked adapter for the peer.
lmcache/v1/multiprocess/modules/p2p_controller.py:493
↓ 1 callersFunction_per_layer_paged_shape
Return the logical shape of a single per-layer paged buffer tensor. Args: engine_kv_format: The format enum that describes how K/V tokens
lmcache/python_ops_fallback.py:815
↓ 1 callersFunction_pick_layout_probe_tensor
()
lmcache/v1/gpu_connector/utils.py:463
↓ 1 callersMethod_poll_for_completion
Poll for sync completion status and resend missing batches if needed. Args: sync_id: The sync session ID key
lmcache/v1/cache_controller/full_sync_sender.py:341
↓ 1 callersMethod_poll_load_results
Query pending load results from signaled adapters.
lmcache/v1/distributed/storage_controllers/prefetch_controller.py:1046
↓ 1 callersMethod_poll_lookup_results
Query pending lookup-and-lock results from signaled adapters.
lmcache/v1/distributed/storage_controllers/prefetch_controller.py:1028
↓ 1 callersMethod_poll_prefix_leg
Poll the CB prefix handle; on completion close the cb.prefix_lookup span. Consume-once: publishes CB_PREFIX_LOOKUP_END exactly once when the
lmcache/v1/multiprocess/modules/blend_v3.py:827
↓ 1 callersMethod_prefix_hash
( self, token_chunks: Iterable[Union[torch.Tensor, List[int]]], )
lmcache/v1/token_database.py:357
↓ 1 callersFunction_print_section
(title: str, data: dict, depth: int = 0)
lmcache/tools/mp_status_viewer/__main__.py:36
↓ 1 callersFunction_print_summary
Print a formatted configuration summary.
lmcache/cli/commands/bench/engine_bench/interactive/__init__.py:107
↓ 1 callersFunction_probe_cufile_register
Try to actually register a cuFile handle on a real file in the test scratch dir. Returns True iff cuFileHandleRegister succeeds. Importa
tests/v1/utils.py:58
↓ 1 callersMethod_process_block
Update shadow map for a single physical block.
lmcache/v1/mp_observability/subscribers/metrics/l0_lifecycle.py:189
↓ 1 callersMethod_process_buffer_with_metadata
Process buffer that contains metadata + data. Used when save_chunk_meta=True (metadata stored remotely).
lmcache/v1/storage_backend/connector/mooncakestore_connector.py:644
↓ 1 callersMethod_process_cached_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector.py:1221
↓ 1 callersMethod_process_cached_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0201.py:1075
↓ 1 callersMethod_process_cached_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0180.py:1012
↓ 1 callersMethod_process_new_keys
Process a batch of newly written keys. 1. Ask the policy which adapters each key should go to. 2. For each adapter target, r
lmcache/v1/distributed/storage_controllers/store_controller.py:556
↓ 1 callersMethod_process_new_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector.py:1200
↓ 1 callersMethod_process_new_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0201.py:1056
↓ 1 callersMethod_process_new_requests
( self, scheduler_output: SchedulerOutput, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0180.py:993
↓ 1 callersMethod_process_qkv
Process QKV tensors. Model-specific implementation.
lmcache/v1/compute/models/base.py:63
↓ 1 callersMethod_process_record
Process a single BlockAllocationRecord.
lmcache/v1/mp_observability/subscribers/metrics/l0_lifecycle.py:163
↓ 1 callersMethod_process_register_response
Process RegisterRetMsg to extract heartbeat_url
lmcache/tools/controller_benchmark/benchmark.py:310
↓ 1 callersMethod_process_register_response
Process RegisterRetMsg and initialize components based on extra_config.
lmcache/v1/cache_controller/worker.py:230
↓ 1 callersMethod_process_retrieve_requests
( self, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector.py:1182
↓ 1 callersMethod_process_retrieve_requests
( self, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0201.py:1038
↓ 1 callersMethod_process_retrieve_requests
( self, metadata: LMCacheMPConnectorMetadata, )
lmcache/integration/vllm/lmcache_mp_connector_0180.py:975
↓ 1 callersFunction_prompt_action
Ask the user to start the benchmark or export config. Returns ``"start"`` or ``"export"``.
lmcache/cli/commands/bench/engine_bench/interactive/__init__.py:124
↓ 1 callersFunction_prompt_choice_fallback
Numbered fallback for non-TTY environments.
lmcache/cli/commands/bench/engine_bench/interactive/terminal.py:273
↓ 1 callersMethod_prune_expired_cache_entries_locked
(self, now: float)
lmcache/v1/distributed/l2_adapters/hfbucket_l2_adapter.py:883
↓ 1 callersMethod_prune_expired_cache_entries_locked
Remove expired cache entries while holding ``_metadata_cache_lock``.
lmcache/v1/storage_backend/connector/hfbucket_connector.py:700
↓ 1 callersMethod_publish_fingerprints
Publish this stored range's chunk fingerprints to the coordinator. Best-effort and fire-and-forget (enqueue only): one wire ``ChunkFi
lmcache/v1/multiprocess/modules/blend_v3.py:1204
↓ 1 callersMethod_publish_skipped
Publish a below-watermark loop tick (no eviction this cycle).
lmcache/v1/distributed/storage_controllers/eviction_controller.py:119
↓ 1 callersMethod_publish_triggered
Publish an above-watermark loop tick (eviction policy ran).
lmcache/v1/distributed/storage_controllers/eviction_controller.py:132
↓ 1 callersMethod_put
(self, key: CacheEngineKey, memory_obj: MemoryObj)
lmcache/v1/storage_backend/connector/redis_connector.py:363
↓ 1 callersMethod_put
(self, key: CacheEngineKey, memory_obj: MemoryObj)
lmcache/v1/storage_backend/connector/redis_connector.py:742
↓ 1 callersMethod_put_and_get_check
(self)
lmcache/v1/health_monitor/checks/remote_backend_check.py:256
↓ 1 callersMethod_put_internal
(self, key: CacheEngineKey, memory_obj: MemoryObj)
lmcache/v1/storage_backend/connector/bigtable_connector.py:352
↓ 1 callersMethod_put_request
(self, key_str: str, mem_obj: MemoryObj)
lmcache/v1/distributed/l2_adapters/s3_l2_adapter.py:899
↓ 1 callersMethod_put_sync
(self, key: CacheEngineKey, memory_obj: MemoryObj)
lmcache/v1/storage_backend/connector/eic_connector.py:444
↓ 1 callersMethod_put_without_metadata
Zero-copy put using put_from when metadata is not stored remotely. This is used when save_chunk_meta=False (matches _batch_get_into).
lmcache/v1/storage_backend/connector/mooncakestore_connector.py:754
↓ 1 callersFunction_python_pipeline
Pure-Python fold -> highest_set_bit -> unfold (no native ops).
benchmarks/microbenchmark/bitmap_ops_benchmark.py:26
↓ 1 callersMethod_query_with_fallback
Send one query and fallback between completions/chat endpoints.
lmcache/cli/commands/query/_request.py:308
↓ 1 callersMethod_queue_item
Add item to async queue with timeout handling.
lmcache/v1/lookup_client/record_strategies/base.py:250
↓ 1 callersMethod_ramp_up
(self, timestamp: float, ramp_up_time: float)
benchmarks/multi_round_qa/multi-round-qa.py:403
↓ 1 callersMethod_read
(self, fd: int, buffer: memoryview, length: int, file_offset: int)
lmcache/v1/storage_backend/connector/hf3fs_connector.py:459
↓ 1 callersMethod_read_bytes_from_offsets
Read bytes from shared memory offsets, skipping initial bytes.
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:523
↓ 1 callersMethod_read_from_config
Read NUMA mapping from the LMCache configuration.
lmcache/v1/system_detection.py:74
↓ 1 callersMethod_read_from_shared_memory
Read data from shared memory using lease offsets. Data format: [RemoteMetadata header (28 bytes)] + [KV cache payload] Data
lmcache/v1/storage_backend/connector/sagemaker_hyperpod_connector.py:417
↓ 1 callersMethod_read_from_sys
Read NUMA mapping from system configuration.
lmcache/v1/system_detection.py:94
↓ 1 callersMethod_read_header
(self)
lmcache/v1/mp_observability/trace/reader.py:110
↓ 1 callersFunction_read_json
(url: str, timeout: float)
lmcache/cli/commands/query/_request.py:86
↓ 1 callersFunction_read_key
Read a single keypress from stdin in raw mode. Returns: ``"up"``, ``"down"``, ``"enter"``, or the literal character.
lmcache/cli/commands/bench/engine_bench/interactive/terminal.py:152
↓ 1 callersMethod_read_meta_header
Read and validate a metadata checkpoint header.
lmcache/v1/storage_backend/raw_block/core.py:1505
↓ 1 callersMethod_read_metadata_info
(self, filename: str)
lmcache/v1/storage_backend/gds_backend.py:423
↓ 1 callersFunction_read_points_by_attrs
``{metric: {sorted_attrs_tuple: value}}`` from the shared reader.
tests/v1/distributed/test_l1_l2_state_metrics.py:130
↓ 1 callersMethod_read_slot_header
Read and decode the slot header at a raw-device offset.
lmcache/v1/storage_backend/raw_block/core.py:1418
↓ 1 callersFunction_read_sysfs_int
Read an integer value from sysfs and return None on failure.
lmcache/v1/storage_backend/raw_block/core.py:120
↓ 1 callersMethod_read_uring_cmd_buffers
Read buffers as bounded NVMe raw-command chunks. Args: offsets: Device offsets for each logical read. buffers: Destin
lmcache/v1/storage_backend/raw_block/core.py:1190
↓ 1 callersFunction_recalculate_blocks_to_skip
Re-calculate the number of blocks to skip for a batch of chunks based on the blocks per chunk and blocks per sliding window WHEN the window si
lmcache/v1/multiprocess/modules/lmcache_driven_transfer.py:203
↓ 1 callersMethod_reconcile
Bring the local peer adapters in line with ``upstream``. Args: upstream: The peers that should each have a live adapter.
lmcache/v1/multiprocess/modules/p2p_controller.py:458
↓ 1 callersMethod_reconcile_failed_store
( self, indexed: list[tuple[ObjectKey, str, int]], )
lmcache/v1/distributed/l2_adapters/hfbucket_l2_adapter.py:834
↓ 1 callersMethod_recover_checkpoint_dtype
Recover checkpoint dtype from entry metadata or legacy key strings. Args: encoded_key: Encoded raw-block key from the checkpoint
lmcache/v1/storage_backend/raw_block/core.py:1835
← previousnext →3,501–3,600 of 13,371, ranked by callers