EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Cherry Pick from pr#3981][0.11.0][P/D]Make kv-transfer env variable take effect & Fix load-balance proxy (#3983)

Browse Source 
### What this PR does / why we need it?
Make kv-transfer env variable take effect & Fix load-balance proxy.
Cherry Pick from #3981

---------
Signed-off-by: nwpu-zxr <zhouxuerong2@huawei.com>
This commit is contained in:
zxr2333
2025-11-08 13:52:33 +08:00
committed by GitHub
parent 8e72758645
commit d4e2a44307
7 changed files with 35 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/ut/kv_connector/test_mooncake_connector.py
View File

@@ -1018,7 +1018,9 @@ class TestMooncakeConnectorWorker(unittest.TestCase):
self.mock_transfer_engine.register_memory.return_value = 0
self.patches = [
patch('os.getenv', return_value="10,11"),
patch(
'vllm_ascend.distributed.mooncake_layerwise_connector.envs_ascend.PHYSICAL_DEVICES',
'10,11'),
patch('torch.Tensor.size', return_value=(10, 16, 8, 16)),
patch('torch.Tensor.element_size', return_value=4),
patch('torch.Tensor.data_ptr', return_value=0x1000),
@@ -1047,8 +1049,6 @@ class TestMooncakeConnectorWorker(unittest.TestCase):
MagicMock()),
patch('vllm_ascend.distributed.mooncake_connector.threading.Event',
MagicMock()),
patch.dict('sys.modules',
{'vllm_ascend.envs': self.envs_ascend_mock}),
]
for p in self.patches:
@@ -1136,4 +1136,4 @@ class TestMooncakeConnectorWorker(unittest.TestCase):
if __name__ == '__main__':
unittest.main()
unittest.main()

8
tests/ut/kv_connector/test_mooncake_layerwise_connector.py
View File

@@ -800,7 +800,9 @@ class TestMooncakeLayerwiseConnectorWorker(unittest.TestCase):
self.mock_transfer_engine.register_memory.return_value = 0
self.patches = [
patch('os.getenv', return_value="10,11"),
patch(
'vllm_ascend.distributed.mooncake_layerwise_connector.envs_ascend.PHYSICAL_DEVICES',
'10,11'),
patch('torch.Tensor.size', return_value=(10, 16, 8, 16)),
patch('torch.Tensor.element_size', return_value=4),
patch('torch.Tensor.data_ptr', return_value=0x1000),
@@ -833,8 +835,6 @@ class TestMooncakeLayerwiseConnectorWorker(unittest.TestCase):
patch(
'vllm_ascend.distributed.mooncake_layerwise_connector.threading.Event',
MagicMock()),
patch.dict('sys.modules',
{'vllm_ascend.envs': self.envs_ascend_mock}),
patch(
'vllm_ascend.distributed.mooncake_layerwise_connector.get_ascend_config',
return_value=SimpleNamespace(pd_tp_ratio=1,
@@ -913,4 +913,4 @@ class TestMooncakeLayerwiseConnectorWorker(unittest.TestCase):
if __name__ == '__main__':
unittest.main()
unittest.main()