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

[Feat]support sequence parallelism by pass for VL models (#5632)

Browse Source
This commit is contained in:
realliujiaxu
2026-02-27 08:27:41 +08:00
committed by GitHub
parent ed175d6d92
commit 5def28dcd3
22 changed files with 460 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/ut/ops/test_mla.py
View File

@@ -157,7 +157,7 @@ class TestAscendMultiHeadLatentAttention(TestBase):
hidden_states = torch.randn(3, self.hidden_size)
mock_forward_context = MagicMock(spec=ForwardContext)
mock_forward_context.sp_enabled = False
mock_forward_context.flash_comm_v1_enabled = False
mock_get_forward_context.return_value = mock_forward_context
mock_mla_forward.return_value = (3, self.hidden_size)

8
tests/ut/spec_decode/test_eagle_proposer.py
View File

@@ -390,7 +390,7 @@ class TestEagleProposerDummyRun(TestBase):
# cpu does not support parallel-group, let alone `sp`
@patch("vllm_ascend.spec_decode.eagle_proposer.get_forward_context",
**{"return_value.sp_enabled": False})
**{"return_value.flash_comm_v1_enabled": False})
@patch("vllm_ascend.spec_decode.eagle_proposer.set_ascend_forward_context")
def test_dummy_run_basic(self, mock_context, mock_get_context):
num_tokens = 32
@@ -406,7 +406,7 @@ class TestEagleProposerDummyRun(TestBase):
# cpu does not support parallel-group, let alone `sp`
@patch("vllm_ascend.spec_decode.eagle_proposer.get_forward_context",
**{"return_value.sp_enabled": False})
**{"return_value.flash_comm_v1_enabled": False})
@patch("vllm_ascend.spec_decode.eagle_proposer.set_ascend_forward_context")
def test_dummy_run_with_prefill(self, mock_context, mock_get_context):
mock_context.return_value.__enter__.return_value = None
@@ -426,7 +426,7 @@ class TestEagleProposerDummyRun(TestBase):
mock_return_context.cudagraph_runtime_mode = CUDAGraphMode.FULL
mock_return_context.capturing = True
# cpu does not support parallel-group, let alone `sp`
mock_return_context.sp_enabled = False
mock_return_context.flash_comm_v1_enabled = False
mock_get_context.return_value = mock_return_context
self.proposer.use_cuda_graph = True
# cpu does not support `torch.ops.vllm.maybe_pad_and_reduce`
@@ -449,7 +449,7 @@ class TestEagleProposerDummyRun(TestBase):
mock_return_context.cudagraph_runtime_mode = CUDAGraphMode.FULL
mock_return_context.capturing = False
# cpu does not support parallel-group, let alone `sp`
mock_return_context.sp_enabled = False
mock_return_context.flash_comm_v1_enabled = False
mock_get_context.return_value = mock_return_context
self.proposer.use_cuda_graph = True
# cpu does not support `torch.ops.vllm.maybe_pad_and_reduce`