xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-13 07:32:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix all commit

Browse Source 
Signed-off-by: inkcherry <mingzhi.liu@amd.com>
This commit is contained in:
inkcherry 2025-11-21 05:05:07 +00:00
parent 795a305b1b
commit 857d93cbfb
4 changed files with 11 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tests/v1/kv_connector/moriio_integration/toy_proxy_server.py
View File

@ -16,12 +16,11 @@ from quart import Quart, make_response, request
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG) logger.setLevel(logging.DEBUG)
prefill_instances = [] prefill_instances: list[dict] = []
decode_instances = [] decode_instances: list[dict] = []
request_nums = 0 request_nums = 0
app = Quart(__name__) app = Quart(__name__)
yield_chunk = set()
IP_PORT_PATTERN = re.compile(r"//(\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}):(\d+)") IP_PORT_PATTERN = re.compile(r"//(\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}):(\d+)")
@ -200,7 +199,10 @@ async def handle_request():
request_nums += 1 request_nums += 1
def extract_ip_port_fast(url): def extract_ip_port_fast(url):
return IP_PORT_PATTERN.search(url).groups() match = IP_PORT_PATTERN.search(url)
if not match:
raise ValueError(f"Invalid URL format: {url}")
return match.groups()
req_data = await request.get_json() req_data = await request.get_json()
request_id = str(uuid.uuid4()) request_id = str(uuid.uuid4())

5
vllm/distributed/kv_transfer/kv_connector/v1/moriio_connector.py
View File

@ -4,7 +4,6 @@ import contextlib
import logging import logging
import math import math
import os import os
import pickle
import queue import queue
import threading import threading
import time import time
@ -1673,7 +1672,7 @@ class MoRIIOConnectorWorker:
) # send local mori io engine meta data ) # send local mori io engine meta data
logger.debug("MoRIIO handshake listener sent metadata") logger.debug("MoRIIO handshake listener sent metadata")
# now we send tensor meta data for each block # now we send tensor meta data for each block
buf = pickle.dumps(layer_name_to_local_kv_cache_metadata) buf = msgpack.dumps(layer_name_to_local_kv_cache_metadata)
sock.send_multipart((identity, b"", buf)) sock.send_multipart((identity, b"", buf))
elif msg == MoRIIOConstants.POP_DONE_RECV: elif msg == MoRIIOConstants.POP_DONE_RECV:
_, req_id = sock.recv_multipart() _, req_id = sock.recv_multipart()
@ -1752,7 +1751,7 @@ class MoRIIOConnectorWorker:
assert 0, f"Unexpected frame! {received_frame = }" assert 0, f"Unexpected frame! {received_frame = }"
buf = received_frame[1] buf = received_frame[1]
self.layer_name_to_remote_kv_cache_metadata[expected_engine_id] = ( self.layer_name_to_remote_kv_cache_metadata[expected_engine_id] = (
pickle.loads(buf) msgpack.loads(buf)
) )
setup_agent_time = time.perf_counter() setup_agent_time = time.perf_counter()

1
vllm/entrypoints/openai/serving_chat.py
View File

@ -328,6 +328,7 @@ class OpenAIServingChat(OpenAIServing):
lora_request=lora_request, lora_request=lora_request,
trace_headers=trace_headers, trace_headers=trace_headers,
priority=request.priority, priority=request.priority,
data_parallel_rank=data_parallel_rank,
) )
generator = self.engine_client.generate( generator = self.engine_client.generate(

3
vllm/entrypoints/openai/serving_engine.py
View File

@ -1172,7 +1172,7 @@ class OpenAIServing:
lora_request: LoRARequest | None, lora_request: LoRARequest | None,
trace_headers: Mapping[str, str] | None, trace_headers: Mapping[str, str] | None,
priority: int, priority: int,
data_parallel_rank: int, data_parallel_rank: int | None,
) -> tuple[EngineCoreRequest, dict[str, Any]]: ) -> tuple[EngineCoreRequest, dict[str, Any]]:
"""Use the Processor to process inputs for AsyncLLM.""" """Use the Processor to process inputs for AsyncLLM."""
tokenization_kwargs: dict[str, Any] = {} tokenization_kwargs: dict[str, Any] = {}
@ -1220,6 +1220,7 @@ class OpenAIServing:
lora_request=lora_request, lora_request=lora_request,
trace_headers=trace_headers, trace_headers=trace_headers,
priority=priority, priority=priority,
data_parallel_rank=None,
) )
generator = self.engine_client.generate( generator = self.engine_client.generate(