[Frontend][RL] Track HTTP weight operation outcomes and concurrency (#55781)

Signed-off-by: Ronald1995 <ronaldautomobile@163.com>
Signed-off-by: aoshen02 <aoshen02@users.noreply.github.com>
Co-authored-by: aoshen02 <aoshen02@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: aoshen02 <aoshen@inferact.ai>
This commit is contained in:
Ronald
2026-09-30 09:19:43 +00:00
committed by GitHub
co-authored by aoshen02 Claude Opus 5.5 aoshen02
parent ae66f9a97c
commit 713ec07655
8 changed files with 476 additions and 14 deletions
@@ -2,7 +2,7 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from http import HTTPStatus
from unittest.mock import AsyncMock, MagicMock
from unittest.mock import MagicMock
import pytest
@@ -29,7 +29,9 @@ LORA_UNLOADING_SUCCESS_MESSAGE = (
)
async def _async_serving_models_init() -> OpenAIServingModels:
async def _async_serving_models_init_with_mock() -> tuple[
OpenAIServingModels, MagicMock
]:
mock_engine_client = MagicMock(spec=EngineClient)
# Set the max_model_len attribute to avoid missing attribute
mock_model_config = MagicMock(spec=ModelConfig)
@@ -45,6 +47,13 @@ async def _async_serving_models_init() -> OpenAIServingModels:
)
await serving_models.init_static_loras()
# The mock is returned separately: `engine_client` is typed as the protocol,
# so assertion helpers are not visible through it.
return serving_models, mock_engine_client
async def _async_serving_models_init() -> OpenAIServingModels:
serving_models, _ = await _async_serving_models_init_with_mock()
return serving_models
@@ -103,7 +112,7 @@ async def test_load_lora_adapter_duplicate():
@pytest.mark.parametrize("load_inplace", [False, True])
async def test_load_lora_adapter_base_model_name(load_inplace: bool):
"""A LoRA named after a served model would shadow it for all requests."""
serving_models = await _async_serving_models_init()
serving_models, mock_engine_client = await _async_serving_models_init_with_mock()
request = LoadLoRAAdapterRequest(
lora_name=MODEL_NAME, lora_path="/path/to/adapter", load_inplace=load_inplace
)
@@ -112,8 +121,7 @@ async def test_load_lora_adapter_base_model_name(load_inplace: bool):
assert response.error.type == "InvalidUserInput"
assert response.error.code == HTTPStatus.BAD_REQUEST
assert len(serving_models.lora_requests) == 0
assert isinstance(serving_models.engine_client.add_lora, AsyncMock)
serving_models.engine_client.add_lora.assert_not_called()
mock_engine_client.add_lora.assert_not_called()
@pytest.mark.asyncio