[Model] Add DeepSeek-V4 CPU backend (#55355)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
@@ -56,7 +56,7 @@ from vllm.v1.attention.backends.mla.flashmla_sparse import (
|
||||
FlashMLASparseMetadataBuilder,
|
||||
triton_convert_req_index_to_global_index,
|
||||
)
|
||||
from vllm.v1.attention.backends.mla.indexer import split_indexer_prefill_chunks
|
||||
from vllm.v1.attention.backends.mla.indexer import DeepseekV32IndexerMetadataBuilder
|
||||
from vllm.v1.attention.backends.utils import (
|
||||
split_decodes_and_prefills,
|
||||
split_prefill_chunks,
|
||||
@@ -1424,7 +1424,7 @@ def test_sparse_backend_prefill_correctness(
|
||||
def test_split_indexer_prefill_chunks(
|
||||
seq_lens, query_lens, workspace_size, max_logits_bytes, expected
|
||||
):
|
||||
out = split_indexer_prefill_chunks(
|
||||
out = DeepseekV32IndexerMetadataBuilder._split_indexer_prefill_chunks(
|
||||
seq_lens,
|
||||
query_lens,
|
||||
workspace_size,
|
||||
@@ -1438,7 +1438,9 @@ def test_split_indexer_prefill_chunks_single_request_overflow():
|
||||
seq_lens = torch.tensor([1000, 50])
|
||||
query_lens = torch.tensor([100, 5])
|
||||
|
||||
out = split_indexer_prefill_chunks(seq_lens, query_lens, 2000, 1000)
|
||||
out = DeepseekV32IndexerMetadataBuilder._split_indexer_prefill_chunks(
|
||||
seq_lens, query_lens, 2000, 1000
|
||||
)
|
||||
# max_logits_elems = 250, N=1000 -> max_q = 1 -> 100 query sub-chunks
|
||||
expected = [(slice(0, 1), slice(i, i + 1)) for i in range(100)]
|
||||
# req1: M=5, N=50 -> 250 elems fits budget
|
||||
|
||||
Reference in New Issue
Block a user