# SPDX-License-Identifier: Apache-2.0 import contextlib import hashlib import math import queue import random import struct import threading import time from collections import defaultdict from collections.abc import Iterator from concurrent.futures import ThreadPoolExecutor from dataclasses import dataclass from typing import TYPE_CHECKING, Any, Callable, List, Optional, Tuple import httpx import msgspec import numpy as np import numpy.typing as npt import torch import zmq from mooncake.engine import TransferEngine # type: ignore from vllm import envs from vllm.config import VllmConfig, get_current_vllm_config from vllm.distributed.kv_transfer.kv_connector.v1.base import ( KVConnectorBase_V1, KVConnectorMetadata, KVConnectorRole) from vllm.distributed.parallel_state import (get_tensor_model_parallel_rank, get_tp_group, get_world_group) from vllm.utils import get_ip, logger, make_zmq_path, make_zmq_socket from vllm.v1.core.sched.output import SchedulerOutput from vllm.v1.request import RequestStatus import vllm_ascend.envs as envs_ascend from vllm_ascend.ascend_config import get_ascend_config from vllm_ascend.distributed.utils import (align_memory, kv_alltoall_and_rearrange) if TYPE_CHECKING: from vllm.attention.backends.abstract import AttentionMetadata from vllm.forward_context import ForwardContext from vllm.v1.core.kv_cache_manager import KVCacheBlocks from vllm.v1.request import Request GET_META_MSG = b"get_meta_msg" DONE_RECVING_MSG = b"done_recving_msg" class MooncakeAgentMetadata(msgspec.Struct, omit_defaults=True, dict=True): engine_id: str te_rpc_port: int kv_caches_base_addr: list[int] num_blocks: int @dataclass class ReqMeta: local_block_ids: list[int] # Not None if layer-wise is disabled remote_block_ids: Optional[list[int]] remote_host: Optional[str] remote_port: Optional[int] remote_engine_id: Optional[str] # Not None if layer-wise is enabled metaserver: Optional[str] remote_tp_size: Optional[int] class DecodeMooncakeAgentMetadata(msgspec.Struct, omit_defaults=True, dict=True): req_id: str block_ids: list[int] host: str port: int engine_id: str te_rpc_port: int kv_caches_base_addr: list[int] num_blocks: int class KVCacheTaskTracker: def __init__(self, target_count: int = 1, on_done: Callable[[str], None] = lambda x: None, on_timeout: Callable[[set[str]], Any] = lambda x: None): super().__init__() self.target_count = target_count self.done_task_lock = threading.Lock() self.done_task_counts: defaultdict[str, int] = defaultdict(int) self.finished_requests: set[str] = set() # Only used in prefill node. Tracks requests whose kv blocks freeing is # intentionally delayed. Each entry is a tuple of (request_id, # timestamp). If a request remains in this queue for too long, it will # be force-freed. # Notice: In layer-wise mode, the transfer may complete before it is # added to delayed_free_requests when prefill node finishes forwarding. # Therefore we need to track requests that are removed before being added. self.delayed_free_requests: dict[str, float] = {} self.removed_delayed_free_requests: set[str] = set() self.on_done = on_done self.on_timeout = on_timeout def update_done_task_count(self, request_id: str): self.done_task_counts[request_id] += 1 if self.done_task_counts[request_id] == self.target_count: with self.done_task_lock: self.finished_requests.add(request_id) self.done_task_counts.pop(request_id) self.on_done(request_id) def get_and_clear_finished_requests(self) -> set[str]: """ Get and clear the requests that have been completed. Returns: A set of request IDs that have been completed. """ with self.done_task_lock: finished_requests = self.finished_requests.copy() expired_requests = self._retrieve_expired_requests() finished_requests.update(expired_requests) self.finished_requests.clear() self.on_timeout(expired_requests) return finished_requests def add_delayed_request(self, request_id: str, delay_start_time: float): """Add a delayed free request, where delay_start_time is monotonic increasing.""" with self.done_task_lock: if request_id in self.removed_delayed_free_requests: self.removed_delayed_free_requests.remove(request_id) else: self.delayed_free_requests[request_id] = delay_start_time def _retrieve_expired_requests(self): """Retrieve all expired delayed requests.""" expired_requests: set[str] = set() # Free delayed requests if they exceed the timeout current_time = time.time() while self.delayed_free_requests: request_id, delay_start_time = next( iter(self.delayed_free_requests.items())) if (current_time - delay_start_time > envs.VLLM_NIXL_ABORT_REQUEST_TIMEOUT): self.delayed_free_requests.pop(request_id) expired_requests.add(request_id) logger.info("Force freed request: %s", request_id) else: break return expired_requests def remove_delayed_request(self, request_id: str): """Remove all delayed free requests matching the given request_id.""" with self.done_task_lock: if self.delayed_free_requests.pop(request_id, None) is None: self.removed_delayed_free_requests.add(request_id) class KVCacheSendingLayerThread(threading.Thread): def __init__(self, tp_rank: int, tp_size: int, decode_tp_size: int, local_engine_id: str, side_channel_host: str, side_channel_port: int, metadata: MooncakeAgentMetadata, ready_event: threading.Event, total_layers: int, engine: TransferEngine, local_kv_base_addr: list[int], block_len: list[int], use_mla: bool, first_kv_cache: torch.Tensor): super().__init__(daemon=True, name="KVCacheSendingLayerThread") self.tp_rank = tp_rank self.tp_size = tp_size self.decode_tp_size = decode_tp_size self.local_engine_id = local_engine_id self.side_channel_host = side_channel_host self.side_channel_port = side_channel_port self.task_tracker = KVCacheTaskTracker(total_layers, on_done=self._post_transfer, on_timeout=self._abort_requests) self.send_layer_thread = SendingLayerThread( self.task_tracker, total_layers, engine, local_kv_base_addr, block_len, use_mla, self.tp_rank, first_kv_cache) self.ready_decode = dict[str, DecodeMooncakeAgentMetadata]() self.pending_decode = dict[str, list[tuple[list[int], int, torch.Tensor, torch.Tensor]]]() self.total_layers = total_layers self.lock = threading.Lock() self.ready_event = ready_event def get_and_clear_finished_requests(self) -> set[str]: """ Get and clear the requests that have been completed. Returns: A set of request IDs that have been completed. """ # vllm won't call us if all inference is done, so we can't do step 9 here return self.task_tracker.get_and_clear_finished_requests() def add_delayed_request(self, request_id: str, delay_start_time: float): return self.task_tracker.add_delayed_request(request_id, delay_start_time) def run(self): """Run the thread to handle KV cache transfer requests.""" self.send_layer_thread.start() handshake_port = self.side_channel_port + self.tp_rank path = make_zmq_path("tcp", self.side_channel_host, handshake_port) logger.info("Starting listening on path: %s", path) with zmq_ctx(zmq.ROUTER, path) as sock: # type: ignore self.ready_event.set() decoder = msgspec.msgpack.Decoder(type=DecodeMooncakeAgentMetadata) while True: try: frames = sock.recv_multipart() if len(frames) < 2: logger.error("Invalid message format: %s", frames) continue identity = frames[0] payload = [f for f in frames[1:] if f != b""] if len(payload) != 1: logger.error("Invalid message format: %s", frames) continue metadata = decoder.decode(payload[0]) request_id = metadata.req_id logger.debug( f"Prefiller has received that request {request_id} from the decoder." ) sock.send_multipart((identity, b"", b"ACK")) self.task_tracker.remove_delayed_request(request_id) with self.lock: self.ready_decode[request_id] = metadata pending = self.pending_decode.pop(request_id, []) for local_block_ids, layer_index, key, value in pending: self.send_layer_thread.send_queue.put( (metadata, request_id, local_block_ids, layer_index, key, value)) except Exception as e: logger.error("Failed to decode message: %s", e) def _post_transfer(self, request_id: str): with self.lock: decoder_meta = self.ready_decode.pop(request_id) path = make_zmq_path("tcp", decoder_meta.host, decoder_meta.port) msg_encoder = msgspec.msgpack.Encoder() encoded_data = msg_encoder.encode(request_id) with zmq_ctx(zmq.REQ, path) as sock: # type: ignore ensure_zmq_send(sock, encoded_data) ack = sock.recv() if ack != b"ACK": raise ValueError(f"Unexpected ACK response: {ack}") def add_request(self, request_id: str, local_block_ids: list[int], layer_index: int, key: torch.Tensor, value: torch.Tensor): # add request to send layer thread with self.lock: if request_id in self.ready_decode: self.send_layer_thread.send_queue.put( (self.ready_decode[request_id], request_id, local_block_ids, layer_index, key, value)) else: self.pending_decode.setdefault(request_id, []).append( (local_block_ids, layer_index, key, value)) def _abort_requests(self, request_ids: set[str]): with self.lock: for request_id in request_ids: self.pending_decode.pop(request_id, None) class SendingLayerThread(threading.Thread): def __init__(self, task_tracker: KVCacheTaskTracker, total_layers: int, engine: TransferEngine, local_kv_base_addr: list[int], block_len: list[int], use_mla: bool, tp_rank: int, first_kv_cache: torch.Tensor): super().__init__(daemon=True, name="KVCacheRecvingPrefillerByeThread") self.send_queue = queue.Queue[tuple[DecodeMooncakeAgentMetadata, str, list[int], int, torch.Tensor, torch.Tensor]]() self.completion_event: Optional[threading.Event] = None self.completion_event_count: int self.task_tracker = task_tracker self.total_layers = total_layers self.local_kv_base_addr = local_kv_base_addr self.block_len = block_len self.use_mla = use_mla self.engine = engine self.tp_rank = tp_rank self.pd_tp_ratio = get_ascend_config().pd_tp_ratio self.num_head_replica = get_ascend_config().num_head_replica self.pd_head_ratio = get_ascend_config().pd_head_ratio vllm_config = get_current_vllm_config() max_model_len = vllm_config.scheduler_config.max_model_len first_kv_cache = first_kv_cache[:max_model_len] alignment = 2 * 1024 * 1024 self.k_buffer = torch.zeros( first_kv_cache.numel() + alignment, dtype=first_kv_cache.dtype, device=first_kv_cache.device) # 【4,1,128】-》【1000， 128】 self.k_buffer = align_memory(self.k_buffer, alignment)[:first_kv_cache.numel()].view( -1, first_kv_cache.shape[-1]) self.v_buffer = torch.zeros(first_kv_cache.numel() + alignment, dtype=first_kv_cache.dtype, device=first_kv_cache.device) self.v_buffer = align_memory(self.v_buffer, alignment)[:first_kv_cache.numel()].view( -1, first_kv_cache.shape[-1]) for tensor in (self.k_buffer, self.v_buffer): assert tensor.data_ptr( ) % alignment == 0, "The address of the registered kv cache should be aligned to 2M" ret_value = self.engine.register_memory(tensor.data_ptr(), tensor.numel()) logger.info( f"Sendinglayerthread register_memory {tensor.data_ptr()} {tensor.numel()} {ret_value=}" ) if ret_value != 0: raise RuntimeError("Mooncake memory registration failed. ") def run(self): """Run the thread to handle KV cache receiving for prefiller bye messages.""" # send kv cache for request in send_queue local_rank = get_world_group().local_rank device = torch.device(f"npu:{local_rank}") torch.npu.set_device(device) while True: request = self.send_queue.get() self._handle_request(request) def _handle_request(self, request: tuple[DecodeMooncakeAgentMetadata, str, list[int], int, torch.Tensor, torch.Tensor]): # send kv layer to remote req_meta, request_id, local_block_ids, layer_index, key, value = request try: logger.debug( f"Starting to transfer KV cache for request {request_id}.") self._transfer_kv_cache(req_meta, local_block_ids, layer_index, key, value) logger.debug( f"Finished transferring KV cache for request {request_id}.") except Exception as e: logger.error("Failed to transfer KV cache for request " f"{request_id}: {e}") finally: self.task_tracker.update_done_task_count(request_id) self.send_queue.task_done() def _transfer_kv_cache(self, req_meta: DecodeMooncakeAgentMetadata, local_block_ids: list[int], layer_index: int, key, value): # send kv layer to remote if len(local_block_ids) == 0: return remote_host = req_meta.host remote_te_port = req_meta.te_rpc_port remote_kv_base_addrs = req_meta.kv_caches_base_addr remote_block_ids = req_meta.block_ids if self.tp_rank % self.num_head_replica != 0: pass elif self.pd_head_ratio == 1: layer_local_kv_base_addr = [ self.local_kv_base_addr[i] for i in [2 * layer_index, 2 * layer_index + 1] ] layer_remote_kv_base_addr = [ remote_kv_base_addrs[i] for i in [2 * layer_index, 2 * layer_index + 1] ] grouped_remote_block_ids, grouped_local_block_ids = \ group_concurrent_contiguous(remote_block_ids, local_block_ids) session_id = f"{remote_host}:{remote_te_port}" src_list, dst_list, length_list = [], [], [] for k, (src_layer_base_addr, dst_layer_base_addr) in enumerate( zip(layer_local_kv_base_addr, layer_remote_kv_base_addr)): block_len = self.block_len[ k % 2] if self.use_mla else self.block_len[0] for group_remote_block_id, group_local_block_id in zip( grouped_remote_block_ids, grouped_local_block_ids): src = src_layer_base_addr + group_local_block_id[ 0] * block_len dst = dst_layer_base_addr + group_remote_block_id[ 0] * block_len length = len(group_local_block_id) * block_len src_list.append(src) dst_list.append(dst) length_list.append(length) torch.npu.synchronize() ret = self.engine.batch_transfer_sync_write( session_id, src_list, dst_list, length_list) if ret < 0: logger.error("Mooncake transfer failed for request %s", req_meta.req_id) raise RuntimeError(f"Mooncake transfer failed, ret: {ret}") else: key = key.view(-1, key.shape[-1]) value = value.view(-1, key.shape[-1]) self.k_buffer[:key.shape[0]].copy_(key) # [:4, 128] -> self.v_buffer[:value.shape[0]].copy_(value) layer_local_kv_base_addr = [ self.k_buffer.data_ptr(), self.v_buffer.data_ptr() ] layer_remote_kv_base_addr = [ remote_kv_base_addrs[i] for i in [2 * layer_index, 2 * layer_index + 1] ] grouped_remote_block_ids, _ = group_concurrent_contiguous( remote_block_ids) session_id = f"{remote_host}:{remote_te_port}" src_list, dst_list, length_list = [], [], [] for k, (src_layer_base_addr, dst_layer_base_addr) in enumerate( zip(layer_local_kv_base_addr, layer_remote_kv_base_addr)): src_layer_addr = src_layer_base_addr for group_remote_block_id in grouped_remote_block_ids: block_len = self.block_len[0] remote_block_len = self.block_len[0] * self.pd_head_ratio src_list.append(src_layer_addr) if src_layer_addr + len( group_remote_block_id ) * block_len > src_layer_base_addr + key.numel( ) * key.element_size(): length = src_layer_base_addr + key.numel( ) * key.element_size() - src_layer_addr else: length = len(group_remote_block_id) * block_len length_list.append(length) dst_list.append(dst_layer_base_addr + group_remote_block_id[0] * remote_block_len + length * ((self.tp_rank // self.num_head_replica) % self.pd_head_ratio)) src_layer_addr += length torch.npu.synchronize() ret = self.engine.batch_transfer_sync_write( session_id, src_list, dst_list, length_list) if ret < 0: logger.error("Mooncake transfer failed for request %s", req_meta.req_id) raise RuntimeError(f"Mooncake transfer failed, ret: {ret}") if self.completion_event is not None: self.completion_event_count -= 1 if self.completion_event_count == 0: self.completion_event.set() self.completion_event = None def add_event(self, event: threading.Event, count: int) -> None: self.completion_event = event self.completion_event_count = count class KVCacheRecvingLayerThread(threading.Thread): def __init__(self, tp_rank: int, side_channel_port: int, tp_size: int, local_engine_id: str, ready_event: threading.Event): super().__init__(daemon=True, name="KVCacheRecvingLayerThread") self.tp_rank = tp_rank self.tp_size = tp_size self.local_engine_id = local_engine_id self.side_channel_host = get_ip() self.side_channel_port = side_channel_port self.lock = threading.Lock() self.done_requests = set[str]() self.ready_event = ready_event def get_and_clear_finished_requests(self) -> set[str]: """ Get and clear the requests that have been completed. Returns: A set of request IDs that have been completed. """ with self.lock: finished_requests = self.done_requests self.done_requests = set() return finished_requests def run(self): """Run the thread to handle KV cache transfer requests.""" #TODO layerwise step9 # with zmq_ctx(zmq.ROUTER, path) as sock: # type: ignore # while True: # recv_msg from prefill request send finish= # Listen for new requests for metadata. # NOTE(rob): we need each rank to have a unique port. This hack to keeps # us moving. We will switch when moving to etcd or where we have a # single ZMQ socket in the scheduler. handshake_port = self.side_channel_port + self.tp_rank path = make_zmq_path("tcp", self.side_channel_host, handshake_port) logger.info("Starting listening on path: %s", path) with zmq_ctx(zmq.ROUTER, path) as sock: # type: ignore self.ready_event.set() decoder = msgspec.msgpack.Decoder(type=str) while True: try: frames = sock.recv_multipart() if len(frames) < 2: logger.error("Invalid message format: %s", frames) continue identity = frames[0] payload = [f for f in frames[1:] if f != b""] if len(payload) != 1: logger.error("Invalid message format: %s", frames) continue request_id = decoder.decode(payload[0]) with self.lock: self.done_requests.add(request_id) sock.send_multipart((identity, b"", b"ACK")) except Exception as e: logger.error("Failed to decode message: %s", e) class MooncakeLayerwiseConnectorMetadata(KVConnectorMetadata): def __init__(self): self.requests: dict[str, ReqMeta] = {} self.requests_to_send: dict[str, float] = {} def add_new_req(self, request_id: str, local_block_ids: list[int], kv_transfer_params: dict[str, Any], metaserver=None): self.requests[request_id] = ReqMeta( local_block_ids=local_block_ids, remote_block_ids=kv_transfer_params.get("remote_block_ids", None), remote_engine_id=kv_transfer_params["remote_engine_id"], remote_host=kv_transfer_params["remote_host"], remote_port=kv_transfer_params["remote_port"], metaserver=metaserver, remote_tp_size=kv_transfer_params.get("remote_tp_size", None), ) class MooncakeLayerwiseConnector(KVConnectorBase_V1): def __init__(self, vllm_config: VllmConfig, role: KVConnectorRole): assert vllm_config.kv_transfer_config is not None self.engine_id = vllm_config.kv_transfer_config.engine_id if role == KVConnectorRole.SCHEDULER: self.connector_scheduler: Optional[MooncakeLayerwiseConnectorScheduler] = \ MooncakeLayerwiseConnectorScheduler(vllm_config, str(self.engine_id)) self.connector_worker: Optional[ MooncakeLayerwiseConnectorWorker] = None elif role == KVConnectorRole.WORKER: self.connector_scheduler = None self.connector_worker = MooncakeLayerwiseConnectorWorker( vllm_config, str(self.engine_id)) ############################################################ # Scheduler Side Methods ############################################################ def get_num_new_matched_tokens( self, request: "Request", num_computed_tokens: int) -> tuple[int, bool]: assert self.connector_scheduler is not None return self.connector_scheduler.get_num_new_matched_tokens( request, num_computed_tokens) def update_state_after_alloc(self, request: "Request", blocks: "KVCacheBlocks", num_external_tokens: int): assert self.connector_scheduler is not None return self.connector_scheduler.update_state_after_alloc( request, blocks, num_external_tokens) def build_connector_meta( self, scheduler_output: SchedulerOutput, ) -> KVConnectorMetadata: assert self.connector_scheduler is not None return self.connector_scheduler.build_connector_meta(scheduler_output) def request_finished( self, request: "Request", block_ids: list[int], ) -> tuple[bool, Optional[dict[str, Any]]]: assert self.connector_scheduler is not None return self.connector_scheduler.request_finished(request, block_ids) def get_finished_count(self) -> Optional[int]: assert self.connector_scheduler is not None return self.connector_scheduler.get_finished_count() ############################################################ # Worker Side Methods ############################################################ def register_kv_caches(self, kv_caches: dict[str, torch.Tensor]): assert self.connector_worker is not None self.connector_worker.register_kv_caches(kv_caches) def get_finished(self, finished_req_ids: set[str]) -> tuple[set[str], set[str]]: """Get the finished recving and sending requests.""" assert self.connector_worker is not None return self.connector_worker.get_finished() def start_load_kv(self, forward_context: "ForwardContext", **kwargs) -> None: assert self.connector_worker is not None assert isinstance(self._connector_metadata, MooncakeLayerwiseConnectorMetadata) self.connector_worker.start_load_kv(self._connector_metadata) def wait_for_layer_load(self, layer_name: str) -> None: """MooncakeLayerwiseConnector does not do layerwise saving.""" assert self.connector_worker is not None assert isinstance(self._connector_metadata, MooncakeLayerwiseConnectorMetadata) self.connector_worker.wait_for_layer_load(layer_name) def save_kv_layer(self, layer_name: str, kv_layer: torch.Tensor, attn_metadata: "AttentionMetadata", **kwargs) -> None: """MooncakeLayerwiseConnector does not save explicitly.""" assert self.connector_worker is not None assert isinstance(self._connector_metadata, MooncakeLayerwiseConnectorMetadata) self.connector_worker.save_kv_layer(layer_name, kv_layer, attn_metadata, self._connector_metadata) def wait_for_save(self): """MooncakeLayerwiseConnector does not save explicitly.""" pass class MooncakeLayerwiseConnectorScheduler: """Implementation of Scheduler side methods""" def __init__(self, vllm_config: VllmConfig, engine_id: str): self.vllm_config = vllm_config self.block_size = vllm_config.cache_config.block_size self.engine_id = engine_id logger.info("Initializing Mooncake Scheduler %s", engine_id) self.side_channel_host = get_ip() self.max_device_id = vllm_config.parallel_config.tensor_parallel_size * \ vllm_config.parallel_config.data_parallel_size # Handshake base port self.side_channel_port = ( vllm_config.kv_transfer_config.kv_port + vllm_config.parallel_config.data_parallel_rank_local * vllm_config.parallel_config.tensor_parallel_size) # Requests that need to start recv. # New requests are added by update_state_after_alloc in # the scheduler. Used to make metadata passed to Worker. self._reqs_need_recv: dict[str, tuple[Request, list[int]]] = {} self._reqs_need_send: dict[str, float] = {} self._reqs_need_send_layerwise: dict[str, tuple[str, int, list[int]]] = {} def get_num_new_matched_tokens( self, request: "Request", num_computed_tokens: int) -> tuple[int, bool]: """ For remote prefill, pull all prompt blocks from remote asynchronously relative to engine execution. Args: request (Request): the request object. num_computed_tokens (int): the number of locally computed tokens for this request Returns: * the number of tokens that can be loaded from the external KV cache beyond what is already computed. * true if the external KV cache tokens will be loaded asynchronously (between scheduler steps). """ params = request.kv_transfer_params logger.debug( "MooncakeLayerwiseConnector get_num_new_matched_tokens: " "num_computed_tokens=%s, kv_transfer_params=%s", num_computed_tokens, params) if params is not None and params.get("do_remote_prefill"): assert num_computed_tokens == 0, "Currently only support " \ "prefill with num_computed_tokens == 0." # Assume that the request's KV cache is already fully prefilled and # can be fetched entirely from the prefill node. count = len(request.prompt_token_ids) if count > 0: return count, True # No remote prefill for this request. return 0, False def update_state_after_alloc(self, request: "Request", blocks: "KVCacheBlocks", num_external_tokens: int): params = request.kv_transfer_params logger.debug( "MooncakeLayerwiseConnector update_state_after_alloc: " "num_external_tokens=%s, kv_transfer_params=%s", num_external_tokens, params) if params is not None and params.get("do_remote_prefill"): if all(p in params for p in ("remote_engine_id", "remote_host", "remote_port")): local_block_ids = (blocks.get_unhashed_block_ids() if num_external_tokens > 0 else []) # Get unhashed blocks to pull from remote. self._reqs_need_recv[request.request_id] = (request, local_block_ids) else: logger.warning( "Got invalid KVTransferParams: %s. This " "request will not utilize KVTransfer", params) params["do_remote_prefill"] = False # Layerwise prefiller add request need send if params is not None and params.get("do_remote_decode"): local_block_ids = (blocks.get_block_ids()[0]) self._reqs_need_send_layerwise[request.request_id] = ( params["metaserver"], len(request.all_token_ids), local_block_ids) def build_connector_meta( self, scheduler_output: SchedulerOutput, ) -> KVConnectorMetadata: meta = MooncakeLayerwiseConnectorMetadata() # Loop through scheduled reqs and convert to ReqMeta. for req_id, (req, block_ids) in self._reqs_need_recv.items(): assert req.kv_transfer_params is not None # For the case where there are no remote blocks to pull # (block_ids is empty), we don't need to schedule # an async read on the worker side. meta.add_new_req( request_id=req_id, local_block_ids=block_ids, kv_transfer_params=req.kv_transfer_params, ) # Clear the list once workers start the transfers self._reqs_need_recv.clear() cached_reqs = scheduler_output.scheduled_cached_reqs new_reqs = scheduler_output.scheduled_new_reqs for req_id, new_blocks in zip(cached_reqs.req_ids, cached_reqs.new_block_ids): if req_id in self._reqs_need_send_layerwise and new_blocks is not None: metaserver, total_tokens, block_ids = self._reqs_need_send_layerwise[ req_id] block_ids.extend(new_blocks[0]) computed_tokens = dict( list(zip(cached_reqs.req_ids, cached_reqs.num_computed_tokens)) + [(x.req_id, x.num_computed_tokens) for x in new_reqs]) for req_id, scheduled_tokens in scheduler_output.num_scheduled_tokens.items( ): if req_id in self._reqs_need_send_layerwise: metaserver, total_tokens, block_ids = self._reqs_need_send_layerwise[ req_id] current_tokens = computed_tokens.get(req_id, 0) + scheduled_tokens if current_tokens == total_tokens: meta.add_new_req( request_id=req_id, local_block_ids=block_ids, kv_transfer_params=defaultdict(lambda: None), metaserver=metaserver) self._reqs_need_send_layerwise.pop(req_id) meta.requests_to_send = self._reqs_need_send self._reqs_need_send = {} return meta def request_finished( self, request: "Request", block_ids: list[int], ) -> tuple[bool, Optional[dict[str, Any]]]: """ Once a request is finished, determine whether request blocks should be freed now or will be sent asynchronously and freed later. """ params = request.kv_transfer_params logger.debug( "MooncakeLayerwiseConnector request_finished, request_status=%s, " "kv_transfer_params=%s", request.status, params) if (params is None or not params.get("do_remote_decode") or request.status != RequestStatus.FINISHED_LENGTH_CAPPED): return False, None computed_block_ids = block_ids delay_free_blocks = len(computed_block_ids) > 0 if delay_free_blocks: logger.info("Delaying free of %d blocks for request %s", len(computed_block_ids), request.request_id) self._reqs_need_send[request.request_id] = time.time() return delay_free_blocks, dict( do_remote_prefill=True, do_remote_decode=False, remote_engine_id=self.engine_id, remote_host=self.side_channel_host, remote_port=self.side_channel_port, remote_block_ids=computed_block_ids, ) def get_finished_count(self) -> Optional[int]: prefill_parallel_config: dict[ str, Any] = self.vllm_config.kv_transfer_config.get_from_extra_config( "prefill", {}) assert "tp_size" in prefill_parallel_config.keys() self._prefill_tp_size = prefill_parallel_config["tp_size"] decode_parallel_config: dict[ str, Any] = self.vllm_config.kv_transfer_config.get_from_extra_config( "decode", {}) assert "tp_size" in decode_parallel_config.keys() self._decode_tp_size = decode_parallel_config["tp_size"] if self.vllm_config.model_config.use_mla: return self._decode_tp_size else: # TODO support mha and gqa return None class MooncakeLayerwiseConnectorWorker: """Implementation of Worker side methods""" def __init__(self, vllm_config: VllmConfig, engine_id: str): self._get_prefill_decode_size(vllm_config) if self._prefill_tp_size < self._decode_tp_size: raise ValueError( f"prefill_tp_size: {self._prefill_tp_size} must be greater than" f" or equal to the decode_tp_size: {self._decode_tp_size}") if TransferEngine is None: raise RuntimeError("mooncake is not available") logger.info("Initializing Mooncake work %s", engine_id) self.engine = TransferEngine() # Metadata. self.completion_event: threading.Event self.vllm_config = vllm_config self.engine_id = engine_id self.tp_rank = get_tensor_model_parallel_rank() self.tp_size = vllm_config.parallel_config.tensor_parallel_size self.tp_group = get_tp_group() self.dp_rank = vllm_config.parallel_config.data_parallel_rank_local self.dp_size = vllm_config.parallel_config.data_parallel_size_local self.kv_caches: dict[str, torch.Tensor] = {} self.side_channel_host = get_ip() self.max_device_id = self.tp_size * self.dp_size self.kv_role = vllm_config.kv_transfer_config.kv_role self.total_layers = vllm_config.model_config.get_num_layers( vllm_config.parallel_config) self.executor = ThreadPoolExecutor(1) self.metaserver_client = httpx.Client( limits=httpx.Limits(max_connections=100000), timeout=None) if self.tp_rank == 0 else None # Handshake base port self.side_channel_port = ( vllm_config.kv_transfer_config.kv_port + vllm_config.parallel_config.data_parallel_rank_local * vllm_config.parallel_config.tensor_parallel_size) self.handshake_port = self.side_channel_port + self.tp_rank self.sockets: dict = {} # get tp device id # TODO(kw): https://github.com/vllm-project/vllm-ascend/pull/940 # introducing some changes device_ids_str = envs_ascend.PHYSICAL_DEVICES if device_ids_str is None: device_ids = list( range(self.dp_rank * self.tp_size, (self.dp_rank + 1) * self.tp_size)) else: device_ids = list(map(int, device_ids_str.split(','))) start_index = self.dp_rank * self.tp_size end_index = start_index + self.tp_size if len(device_ids) < end_index: raise ValueError( f"Not enough physical devices available for DP rank {self.dp_rank}. " f"Expected at least {end_index} devices, but found {len(device_ids)} " "in PHYSICAL_DEVICES.") device_ids = device_ids[start_index:end_index] assert len(device_ids) > self.tp_rank # type: ignore self.device_id = device_ids[self.tp_rank] # type: ignore if vllm_config.kv_transfer_config.get_from_extra_config( 'use_ascend_direct', False): hostname = self.side_channel_host else: hostname = f"{self.side_channel_host}:0:npu_{self.device_id}" self._initialize(hostname=hostname, device_name=None) self.te_rpc_port = self.engine.get_rpc_port() # Background thread for sending or receiving KV caches. self.kv_send_layer_thread: Optional[KVCacheSendingLayerThread] = None self.kv_recv_layer_thread: Optional[KVCacheRecvingLayerThread] = None self.vllm_config = vllm_config self.block_size = vllm_config.cache_config.block_size self.kv_caches_base_addr: list[int] = [] self.pd_tp_ratio = get_ascend_config().pd_tp_ratio self.pd_head_ratio = get_ascend_config().pd_head_ratio self.first_kv_cache = None def _get_prefill_decode_size(self, vllm_config: VllmConfig): # get prefill tp and dp size from extra config prefill_parallel_config: dict[ str, Any] = vllm_config.kv_transfer_config.get_from_extra_config( "prefill", {}) assert "tp_size" in prefill_parallel_config.keys() self._prefill_tp_size = prefill_parallel_config["tp_size"] assert "dp_size" in prefill_parallel_config.keys() self._prefill_dp_size = prefill_parallel_config["dp_size"] # get decode tp and dp size from extra config decode_parallel_config: dict[ str, Any] = vllm_config.kv_transfer_config.get_from_extra_config( "decode", {}) assert "tp_size" in decode_parallel_config.keys() self._decode_tp_size = decode_parallel_config["tp_size"] assert "dp_size" in decode_parallel_config.keys() self._decode_dp_size = decode_parallel_config["dp_size"] def _initialize( self, hostname: str, device_name: Optional[str], ) -> None: """Initialize the mooncake instance.""" device_name = device_name if device_name is not None else "" ret_value = self.engine.initialize(hostname, "P2PHANDSHAKE", "ascend", device_name) if ret_value != 0: raise RuntimeError( f"Mooncake initialization failed with ret_value: {ret_value}") def register_kv_caches(self, kv_caches: dict[str, torch.Tensor]): """Register the KV Cache data.""" _, first_kv_cache_tuple = next(iter(kv_caches.items())) first_kv_cache = first_kv_cache_tuple[0] self.first_kv_cache = first_kv_cache # TODO(tms): Find a more robust way to detect and handle MLA self.use_mla = first_kv_cache_tuple[0].size( -1) != first_kv_cache_tuple[1].size(-1) if self.use_mla: # MLA case.[num_block, block_size, 1, hidden_dim] self.num_blocks = first_kv_cache.shape[0] block_rank = 3 # [block_size, latent_dim] block_shape_norm = first_kv_cache_tuple[0].shape[-block_rank:] block_shape_pe = first_kv_cache_tuple[1].shape[-block_rank:] self.block_len = [ first_kv_cache[0].element_size() * math.prod(block_shape_norm), first_kv_cache[1].element_size() * math.prod(block_shape_pe) ] logger.info( "num_blocks: %s, block_shape_norm: %s, block_shape_pe: %s", self.num_blocks, block_shape_norm, block_shape_pe) else: # [num_block, block_size, num_head, hidden_dim] self.num_blocks = first_kv_cache.shape[0] kv_elem_size = first_kv_cache.element_size() block_rank = 3 # [block_size, kv_heads, head_dim] block_shape = first_kv_cache.shape[-block_rank:] self.block_len = [kv_elem_size * math.prod(block_shape)] logger.info("num_blocks: %s, block_shape: %s", self.num_blocks, block_shape) logger.info("Registering KV_Caches. use_mla: %s, shape %s", self.use_mla, first_kv_cache.shape) self.kv_caches = kv_caches kv_caches_base_addr = [] for cache_or_caches in kv_caches.values(): # Normalize to always be a list of caches if self.use_mla: for i, cache in enumerate(cache_or_caches, 0): base_addr = cache.data_ptr() region_len = self.num_blocks * self.block_len[i % 2] kv_caches_base_addr.append(base_addr) self._register(base_addr, region_len) else: cache_list = [cache_or_caches ] if self.use_mla else cache_or_caches for cache in cache_list: base_addr = cache.data_ptr() region_len = self.num_blocks * self.block_len[0] kv_caches_base_addr.append(base_addr) self._register(base_addr, region_len) self.kv_caches_base_addr = kv_caches_base_addr # After KV Caches registered, start the sending or receiving thread. metadata = MooncakeAgentMetadata( engine_id=self.engine_id, te_rpc_port=self.te_rpc_port, kv_caches_base_addr=kv_caches_base_addr, num_blocks=self.num_blocks, ) ready_event = threading.Event() if self.kv_role == 'kv_producer': self.kv_send_layer_thread = KVCacheSendingLayerThread( self.tp_rank, self.tp_size, self._decode_tp_size, self.engine_id, self.side_channel_host, self.side_channel_port, metadata, ready_event, self.total_layers, self.engine, kv_caches_base_addr, self.block_len, self.use_mla, self.first_kv_cache) self.kv_send_layer_thread.start() else: self.kv_recv_layer_thread = KVCacheRecvingLayerThread( self.tp_rank, self.side_channel_port, self.tp_size, self.engine_id, ready_event) self.kv_recv_layer_thread.start() ready_event.wait() def _register(self, ptr, length): logger.info( "Registering KV cache: ptr=0x%x, length=%d, num_blocks=%d, " "block_lens=%s", ptr, length, self.num_blocks, self.block_len) ret_value = self.engine.register_memory(ptr, length) if ret_value != 0: raise RuntimeError("Mooncake memory registration failed.") def _access_metaserver(self, url, message): self.metaserver_client.post(url, json=message) def get_finished(self) -> tuple[set[str], set[str]]: done_sending = ( self.kv_send_layer_thread. get_and_clear_finished_requests( # type: ignore[union-attr] ) if self.kv_role == 'kv_producer' else set()) done_recving = ( self.kv_recv_layer_thread. get_and_clear_finished_requests( # type: ignore[union-attr] ) if self.kv_role == 'kv_consumer' else set()) if self.tp_rank == 0: logger.debug( "Number of completed KV cache send requests: %d, receive " "requests: %d", len(done_sending), len(done_recving)) return done_sending, done_recving def start_load_kv(self, metadata: MooncakeLayerwiseConnectorMetadata): """Start loading KV blocks from remote engine.""" self.current_layer = 0 if self.vllm_config.kv_transfer_config.is_kv_producer: for req_id, meta in metadata.requests.items(): logger.debug( f"Send request: {req_id} to proxy metaserver: {meta.metaserver}" ) if self.tp_rank == 0: # All parameters here should appear in the returned dict of # request_finished in the scheduler side except "request_id". kv_transfer_params = dict( request_id=req_id, do_remote_prefill=True, do_remote_decode=False, remote_engine_id=self.engine_id, remote_host=self.side_channel_host, remote_port=self.side_channel_port) future = self.executor.submit( self._access_metaserver, url=meta.metaserver, message=kv_transfer_params, ) def handle_exception(future): if future.exception(): logger.error( f"Access metaserver fail: {future.exception()}" ) future.add_done_callback(handle_exception) else: for req_id, meta in metadata.requests.items(): for offset in range(self.pd_tp_ratio): path = make_zmq_path( "tcp", meta.remote_host, meta.remote_port + self.tp_rank * self.pd_tp_ratio + offset) logger.info( f"Notify the prefiller: {path} that request: {req_id} from decoder is ready." ) msg_encoder = msgspec.msgpack.Encoder() docode_metadata = DecodeMooncakeAgentMetadata( req_id=req_id, block_ids=meta.local_block_ids, port=self.handshake_port, host=self.side_channel_host, engine_id=self.engine_id, te_rpc_port=self.te_rpc_port, kv_caches_base_addr=self.kv_caches_base_addr, num_blocks=self.num_blocks) encoded_data = msg_encoder.encode(docode_metadata) size_in_bytes = len(encoded_data) logger.debug( "Size of encoded Mooncake agent metadata: %d bytes", size_in_bytes) with zmq_ctx(zmq.REQ, path) as sock: # type: ignore ensure_zmq_send(sock, encoded_data) ack = sock.recv() if ack != b"ACK": raise ValueError( f"Unexpected ACK from prefill node: {ack}") if self.kv_send_layer_thread is not None: for req_id, delay_start_time in metadata.requests_to_send.items(): if self.tp_rank in self._get_remote_tp_ranks_for_req(req_id): self.kv_send_layer_thread.add_delayed_request( req_id, delay_start_time) def save_kv_layer(self, layer_name: str, kv_layer: Tuple[torch.Tensor, torch.Tensor], attn_metadata: "AttentionMetadata", connector_metadata: MooncakeLayerwiseConnectorMetadata, **kwargs) -> None: """MooncakeLayerwiseConnector does not save explicitly.""" if self.kv_role == 'kv_producer': if self.pd_head_ratio != 1: if self.current_layer != 0: self.completion_event.wait() self.completion_event = threading.Event() if self.kv_send_layer_thread is not None: self.kv_send_layer_thread.send_layer_thread.add_event( self.completion_event, len(connector_metadata.requests.keys())) def sort_kv_cache(input_kv: list[list[int]]): return torch.cat([ torch.chunk(tensor, self.pd_head_ratio, dim=0)[x] for x in range(self.pd_head_ratio) for tensor in input_kv ]) total_block_ids = [ request.local_block_ids for request in connector_metadata.requests.values() ] keys = [ kv_layer[0][block_ids].reshape( -1, *kv_layer[0].shape[2:]).clone() for block_ids in total_block_ids ] values = [ kv_layer[1][block_ids].reshape( -1, *kv_layer[1].shape[2:]).clone() for block_ids in total_block_ids ] key_block_size = keys[0].size(0) // len(total_block_ids[0]) value_block_size = values[0].size(0) // len(total_block_ids[0]) keys = sort_kv_cache(keys) # [req1_key, req2_key] values = sort_kv_cache(values) (keys, values) = kv_alltoall_and_rearrange(self.pd_head_ratio, keys, values) key_start_id = 0 value_start_id = 0 else: key = None value = None for req_id, request in connector_metadata.requests.items(): logger.info(f"Add request {req_id} to kv send layer thread. ") if self.pd_head_ratio != 1: key_block_num = len( request.local_block_ids) * key_block_size value_block_num = len( request.local_block_ids) * value_block_size key = keys[key_start_id:key_start_id + key_block_num] #.clone().contiguous() value = values[value_start_id:value_start_id + value_block_num] #.clone().contiguous() key_start_id += key_block_num value_start_id += value_block_num if self.kv_send_layer_thread is not None: self.kv_send_layer_thread.add_request( request_id=req_id, local_block_ids=request.local_block_ids, layer_index=self.current_layer, key=key, value=value) self.current_layer += 1 def wait_for_layer_load(self, layer_name: str) -> None: pass def _get_remote_tp_rank(self, req_id: str) -> int: return self._get_remote_tp_ranks_for_req(req_id)[self.tp_rank] def _get_remote_tp_ranks_for_req(self, req_id: str) -> list[int]: if self._prefill_tp_size == self._decode_tp_size: return list(range(self._prefill_tp_size)) seed = string_to_int64_hash(req_id) rand = random.Random(seed) sampled_nums = rand.sample(range(self._prefill_tp_size), self._decode_tp_size) return sampled_nums @contextlib.contextmanager def zmq_ctx(socket_type: Any, addr: str) -> Iterator[zmq.Socket]: # type: ignore """Context manager for a ZMQ socket""" if socket_type not in (zmq.ROUTER, zmq.REQ, zmq.DEALER): # type: ignore raise ValueError(f"Unexpected socket type: {socket_type}") ctx: Optional[zmq.Context] = None # type: ignore try: ctx = zmq.Context() # type: ignore yield make_zmq_socket(ctx=ctx, path=addr, socket_type=socket_type, bind=socket_type == zmq.ROUTER) # type: ignore finally: if ctx is not None: ctx.destroy(linger=0) def group_concurrent_contiguous( src: List[int], dst: List[int] = [] ) -> Tuple[List[npt.NDArray[np.int64]], List[npt.NDArray[np.int64]]]: """Vectorised NumPy implementation.""" if not dst: src_only_indices: npt.NDArray[np.int64] = np.array(src, dtype=np.int64) if src_only_indices.size == 0: return [], [] brk = np.where((np.diff(src_only_indices) != 1))[0] + 1 src_groups = np.split(src_only_indices, brk) src_groups = [g.tolist() for g in src_groups] return src_groups, [] else: src_indices: npt.NDArray[np.int64] = np.array(src, dtype=np.int64) dst_indices: npt.NDArray[np.int64] = np.array(dst, dtype=np.int64) if src_indices.size == 0: return [], [] brk = np.where((np.diff(src_indices) != 1) | (np.diff(dst_indices) != 1))[0] + 1 src_groups = np.split(src_indices, brk) dst_groups = np.split(dst_indices, brk) src_groups = [g.tolist() for g in src_groups] dst_groups = [g.tolist() for g in dst_groups] return src_groups, dst_groups def string_to_int64_hash(input_str): """ Hash the string using SHA-256 and convert it into an int64 integer. """ hashed_bytes = hashlib.sha256(input_str.encode("utf-8")).digest() trunked_bytes = hashed_bytes[:8] uint64_value = struct.unpack(" 0: logger.warning( f"Send failed: {e}, retrying... ({retries_left} " "attempts left)") time.sleep(0.1) else: logger.error(f"Send failed after all retries: {e}") raise RuntimeError(f"Failed to send data after {max_retries} " f"retries: {e}") def ensure_zmq_recv( socket: zmq.Socket, # type: ignore poller: zmq.Poller, # type: ignore timeout: float = 1.0, max_retries: int = 3) -> bytes: retries_left = max_retries while True: try: if dict(poller.poll(int(timeout * 1000))): # milliseconds data = socket.recv() return data else: raise zmq.ZMQError("Receive timeout") # type: ignore except zmq.ZMQError as e: # type: ignore retries_left -= 1 if retries_left > 0: logger.warning(f"Receive failed: {e}, retrying... " f"({retries_left} attempts left)") time.sleep(0.1) else: logger.error(f"Receive failed after all retries: {e}") raise RuntimeError( f"Failed to receive data after {max_retries} " f"retries: {e}")