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 9, 18:15 UTC
[ROCm][PD] Remove unused moriio connector proxy code (#32939)
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Mark McLoughlin committed
Jan 23, 2026 at 15:59 UTC
1cb4341fbcca4e58fa1ac5c1d55691eff16bf52b
1 file changed
-22
examples/online_serving/disaggregated_serving/moriio_toy_proxy_server.py
-22
@@ -166,27 +166,6 @@ async def stream_decode_response(session, response, request_id):
166
await session.close()
167
168
169
-async def send_request_to_decode(endpoint, req_data, request_id):
170
- async with aiohttp.ClientSession(
171
- timeout=aiohttp.ClientTimeout(total=6 * 6000 * 6000)
172
- ) as session:
173
- headers = {
174
- "Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}",
175
- "X-Request-Id": request_id,
176
- }
177
- async with session.post(
178
- url=endpoint, json=req_data, headers=headers
179
- ) as response:
180
- if response.status == 200:
181
- async for chunk_bytes in response.content.iter_chunked(1024):
182
- yield chunk_bytes
183
- else:
184
- raise RuntimeError(
185
- "send_request_to_decode response.status != 200,response.statuus = ",
186
- response.status,
187
- )
188
-
189
-
169
def example_round_robin_dp_loader(request_number, dp_size):
170
return request_nums % dp_size
171
@@ -233,7 +212,6 @@ async def handle_request():
212
)
213
214
dip, dport = extract_ip_port_fast(decode_instance_endpoint["request_address"])
236
- ip, port = extract_ip_port_fast(prefill_instance_endpoint["request_address"])
215
216
req_data_to_prefill = copy.deepcopy(req_data)
217
req_data_to_prefill["kv_transfer_params"] = {}