EngineX-Ascend/enginex-ascend-910-vllm
10
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init v0.11.0rc0

Browse Source
This commit is contained in:
luopingyi
2025-10-14 10:38:28 +08:00
parent 67afd0ea78
commit 66dc16f966
278 changed files with 28130 additions and 11708 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
tests/ut/torchair/test_torchair_mla.py
View File

@@ -190,12 +190,15 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.scheduler_config.chunked_prefill_enabled = False
mock_device = 'cpu'
mock_vllm_config.speculative_config = None
ascend_config = MagicMock()
ascend_config.torchair_graph_config = MagicMock()
ascend_config.torchair_graph_config.enabled = True
with patch("vllm_ascend.torchair.torchair_mla.get_ascend_config",
return_value=ascend_config):
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
self.assertEqual(builder.block_size,
@@ -216,7 +219,10 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
ascend_config.torchair_graph_config = MagicMock()
ascend_config.torchair_graph_config.enabled = True
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
input_batch = MagicMock()
@@ -250,9 +256,12 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.scheduler_config.chunked_prefill_enabled = False
mock_device = 'cpu'
mock_vllm_config.speculative_config = None
with patch("vllm_ascend.torchair.torchair_mla.get_ascend_config",
return_value=ascend_config):
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
input_batch = MagicMock()
@@ -285,7 +294,10 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.scheduler_config.chunked_prefill_enabled = False
mock_device = 'cpu'
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
block_tables = torch.randint(0, 100, (3, 10), dtype=torch.int32)
@@ -305,7 +317,10 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.scheduler_config.chunked_prefill_enabled = False
mock_device = 'cpu'
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
block_tables = torch.randint(0, 100, (3, 10), dtype=torch.int32)
@@ -326,7 +341,10 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.scheduler_config.chunked_prefill_enabled = False
mock_device = 'cpu'
builder = AscendMLATorchairMetadataBuilder(mock_vllm_config,
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(None, None,
mock_vllm_config,
mock_device)
block_tables = torch.randint(0, 100, (3, 10), dtype=torch.int32)
@@ -351,7 +369,11 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
mock_vllm_config.model_config.dtype = torch.float16
mock_device = 'cpu'
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(
None,
None,
mock_vllm_config,
mock_device,
metadata_cls=AscendMLATorchairMetadata)
@@ -416,7 +438,11 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
model = MagicMock(spec=nn.Module)
model.model = MagicMock(spec=nn.Module)
mock_vllm_config.speculative_config = None
builder = AscendMLATorchairMetadataBuilder(
None,
None,
mock_vllm_config,
mock_device,
metadata_cls=AscendMLATorchairMetadata)
@@ -437,14 +463,16 @@ class TestAscendMLATorchairMetadataBuilder(TestBase):
max_query_len=1,
decode_token_per_req=torch.tensor([1, 1, 1]),
block_table_tensor=torch.zeros((10, 10)),
slot_mapping_cpu=torch.tensor(range(20)),
slot_mapping=torch.tensor(range(20)),
actual_seq_lengths_q=torch.tensor([0, 1, 2]),
positions=torch.tensor([1, 1]),
attn_mask=torch.ones((15, 15)),
spec_attn_mask=None,
attn_state=AscendAttentionState.ChunkedPrefill)
attn_state=AscendAttentionState.ChunkedPrefill,
num_computed_tokens_cpu=None,
seq_lens=None)
metadata = builder.build(common_attn_metadata, model)
metadata = builder.build(1, common_attn_metadata, model)
self.assertIsInstance(metadata, AscendMLATorchairMetadata)
self.assertEqual(metadata.num_input_tokens, 0)