A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 4, 13:45 UTC
[CI][AMD][BugFix] Update request URL in test_moriio_connector to match vllm-router compatibility changes (#41076)
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
rasmith committed
Apr 28, 2026 at 13:14 UTC
358a755e43b07b9454904df9d3c3fae3340058f1
1 file changed
+18
-8
tests/v1/kv_connector/unit/test_moriio_connector.py
+18
-8
@@ -3,6 +3,7 @@
3
import importlib.util
4
import os
5
import subprocess
6
+import uuid
7
from unittest.mock import MagicMock, patch
8
9
import msgspec
@@ -99,6 +100,11 @@ def _setup_kv_transfer_request(
100
"remote_engine_id": "test_engine",
101
}
102
)
103
+ zmq_addr = f"host:{remote_host},handshake:{fake_port},notify:{fake_port}"
104
+ fake_uuid = uuid.uuid4().hex
105
+ request.request_id = (
106
+ f"___prefill_addr_{zmq_addr}___decode_addr_{zmq_addr}_{fake_uuid}"
107
+ )
108
return request
109
110
@@ -254,13 +260,14 @@ def test_write_mode_saves_local_block_ids():
260
do_remote_decode=True,
261
do_remote_prefill=False,
262
)
263
+
264
+ # Setup KV transfer params and embed ZMQ addrs in request_id before
265
+ # adding to scheduler so the ID is consistent everywhere.
266
+ request = _setup_kv_transfer_request(request)
267
request_id = request.request_id
268
269
scheduler.add_request(request)
270
261
- # Fake Config
262
- request = _setup_kv_transfer_request(request)
263
-
271
# Remote Prefill, triggers MoRIIOConnectorMetadata.
272
scheduler_output = scheduler.schedule()
273
kv_connector_metadata = scheduler_output.kv_connector_metadata
@@ -312,13 +319,14 @@ def test_write_mode_with_chunked_prefill_saves_local_block_ids():
319
do_remote_decode=True,
320
do_remote_prefill=False,
321
)
322
+
323
+ # Setup KV transfer params and embed ZMQ addrs in request_id before
324
+ # adding to scheduler so the ID is consistent everywhere.
325
+ request = _setup_kv_transfer_request(request)
326
request_id = request.request_id
327
328
scheduler.add_request(request)
329
319
- # Fake Config
320
- request = _setup_kv_transfer_request(request)
321
-
330
# Remote Prefill with chunked prefill, triggers multiple schedules.
331
expected_counts = [(0, 0, 0), (0, 0, 0), (1, 0, 0)]
332
kv_connector_metadata = None
@@ -363,6 +371,10 @@ def test_read_mode_loads_remote_block_ids(moriio_read_mode):
371
do_remote_decode=False,
372
do_remote_prefill=True,
373
)
374
+
375
+ # Setup KV transfer params and embed ZMQ addrs in request_id before
376
+ # adding to scheduler so the ID is consistent everywhere.
377
+ request = _setup_kv_transfer_request(request)
378
request_id = request.request_id
379
380
scheduler.add_request(request)
@@ -370,8 +382,6 @@ def test_read_mode_loads_remote_block_ids(moriio_read_mode):
382
0
383
].req_to_blocks[request_id]
384
373
- request = _setup_kv_transfer_request(request)
374
-
385
# Set remote block ids to be fetched.
386
request.kv_transfer_params["remote_block_ids"] = block_list
387