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

[bugfix][refactor] fix recompute_scheduler break with vllm 0.12.0 & support async scheduling & refactor recompute_scheduler.py (#4895)

Browse Source 
### What this PR does / why we need it?
Currently, the initialization and fundamental functions of
RecomputeScheduler are broken with `vLLM v0.12.0`. This PR fixes the
conflicts of `RecomputeScheduler` and refactor its implementations by
inheriting original `Scheduler` of vLLM. Meanwhile, this PR also
supports async cheduling with recompute scheduler by implementing
`AsyncRecomputeScheduler` which is simply inherited `AsncyScheduler` of
vLLM and `RecomputeScheduler` of vLLM-Ascend with python MRO.
### Does this PR introduce _any_ user-facing change?
No. The switch naming is the same as v0.11.0 :
`recompute_scheduler_enable`
### How was this patch tested?
E2E serving with 2P1D dsv3.1 passed. The performance was the same as
original vllm scheduler with `async_scheduling` and preempted requests
in D Nodes are successfully transfered to Proxy and further to P Node.
This significantly improves the performance and robustness of PD
disaggregation deployments.


- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: linfeng-yuan <1102311262@qq.com>
This commit is contained in:
linfeng-yuan
2025-12-11 22:24:49 +08:00
committed by GitHub
parent e538fa6f9c
commit 0fbe0831ec
4 changed files with 320 additions and 911 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
tests/ut/test_platform.py
View File

@@ -234,7 +234,7 @@ class TestNPUPlatform(TestBase):
return_value=AscendDeviceType._910_93)
@patch("os.environ", {})
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_basic_config_update(
self, mock_init_recompute, mock_soc_version, mock_update_acl,
@@ -266,7 +266,7 @@ class TestNPUPlatform(TestBase):
return_value=AscendDeviceType._910_93)
@patch("vllm_ascend.ascend_config.init_ascend_config")
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_no_model_config_warning(
self, mock_init_recompute, mock_init_ascend, mock_soc_version):
@@ -291,7 +291,7 @@ class TestNPUPlatform(TestBase):
return_value=AscendDeviceType._910_93)
@patch("vllm_ascend.ascend_config.init_ascend_config")
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_enforce_eager_mode(
self, mock_init_recompute, mock_init_ascend, mock_soc_version):
@@ -328,7 +328,7 @@ class TestNPUPlatform(TestBase):
@patch("vllm_ascend.utils.update_default_aclgraph_sizes")
@patch("vllm_ascend.ascend_config.init_ascend_config")
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_unsupported_compilation_level(
self, mock_init_recompute, mock_init_ascend, mock_update_default,
@@ -397,7 +397,7 @@ class TestNPUPlatform(TestBase):
return_value=AscendDeviceType._910_93)
@patch("vllm_ascend.ascend_config.init_ascend_config")
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_cache_config_block_size(
self, mock_init_recompute, mock_init_ascend, mock_soc_version):
@@ -424,7 +424,7 @@ class TestNPUPlatform(TestBase):
return_value=AscendDeviceType._910_93)
@patch("vllm_ascend.ascend_config.init_ascend_config")
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_v1_worker_class_selection(
self, mock_init_recompute, mock_init_ascend, mock_soc_version):
@@ -462,7 +462,7 @@ class TestNPUPlatform(TestBase):
@patch('vllm_ascend.utils.get_ascend_device_type',
return_value=AscendDeviceType._310P)
@patch(
"vllm_ascend.core.recompute_schedule_config.RecomputeSchedulerConfig.initialize_from_config"
"vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
)
def test_check_and_update_config_310p_no_custom_ops(
self, mock_init_recompute, mock_soc_version, mock_init_ascend):