frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 23:03:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Tests] fix initialization of kv hash in tests (#24273)

Browse Source 
Signed-off-by: Mickael Seznec <mickael@mistral.ai>
This commit is contained in:
Mickaël Seznec
2025-09-15 23:48:27 +02:00
committed by GitHub
parent fd2f10546c
commit 45bfa49cb8
2 changed files with 20 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
tests/v1/core/test_kv_cache_utils.py
View File

@ -30,6 +30,16 @@ from vllm.v1.request import Request
# yapf: enable
@pytest.fixture(autouse=True)
def _auto_init_hash_fn(request):
hash_fn: Callable
if "hash_fn" in request.fixturenames:
hash_fn = init_none_hash(request.getfixturevalue("hash_fn"))
else:
hash_fn = sha256
init_none_hash(hash_fn)
def make_request(
request_id: str,
prompt_token_ids: list[int],
@ -424,7 +434,6 @@ def test_generate_block_hash_extra_keys_cache_salt():
@pytest.mark.parametrize("hash_fn", [sha256, sha256_cbor])
def test_hash_block_tokens(hash_fn):
init_none_hash(hash_fn)
parent_block_hash = BlockHash(b"123")
curr_block_token_ids = (1, 2, 3)
extra_keys = ("key1", "key2")
@ -437,8 +446,6 @@ def test_hash_block_tokens(hash_fn):
@pytest.mark.parametrize("hash_fn", [sha256, sha256_cbor])
def test_request_block_hasher(hash_fn):
kv_cache_utils.init_none_hash(hash_fn)
request = make_request(
request_id="0",
prompt_token_ids=[_ for _ in range(6)],
@ -461,8 +468,6 @@ def test_request_block_hasher(hash_fn):
@pytest.mark.parametrize("hash_fn", [sha256, sha256_cbor])
def test_hash_tokens_different_mm_input(hash_fn):
init_none_hash(hash_fn)
request1 = make_request(
request_id="0",
prompt_token_ids=[_ for _ in range(6)],
@ -491,8 +496,6 @@ def test_hash_tokens_different_mm_input(hash_fn):
@pytest.mark.parametrize("hash_fn", [sha256, sha256_cbor])
def test_hash_request_tokens_no_mm_inputs(hash_fn):
kv_cache_utils.init_none_hash(hash_fn)
request = make_request(
request_id="0",
prompt_token_ids=[_ for _ in range(6)],

15
tests/v1/core/test_prefix_caching.py
View File

@ -25,6 +25,16 @@ from vllm.v1.kv_cache_interface import (FullAttentionSpec, KVCacheConfig,
KVCacheGroupSpec, SlidingWindowSpec)
@pytest.fixture(autouse=True)
def _auto_init_hash_fn(request):
hash_fn: Callable
if "hash_fn" in request.fixturenames:
hash_fn = init_none_hash(request.getfixturevalue("hash_fn"))
else:
hash_fn = sha256
init_none_hash(hash_fn)
def make_request(
request_id: str,
prompt_token_ids: list[int],
@ -105,7 +115,6 @@ def make_kv_cache_config_hybrid_model(block_size: int,
@pytest.mark.parametrize("hash_fn", [sha256, sha256_cbor])
def test_prefill(hash_fn):
init_none_hash(hash_fn)
block_size = 16
manager = KVCacheManager(
@ -736,7 +745,6 @@ def test_cache_blocks(hash_fn):
This is a unit test that tests the correctness of the _cache_full_blocks
function of KVCacheManager.
"""
init_none_hash(hash_fn)
block_size = 4
block_pool = BlockPool(
@ -849,7 +857,6 @@ def test_mm_prefix_caching():
"""
This tests that the multi-modal prefix caching is correct.
"""
kv_cache_utils.init_none_hash(sha256)
block_size = 16
manager = KVCacheManager(
@ -942,8 +949,6 @@ def test_cache_key_salting():
This tests that cache salts are applied during hashing and the cache
is separated cache as expected.
"""
kv_cache_utils.init_none_hash(sha256)
block_size = 16
manager = KVCacheManager(
make_kv_cache_config(block_size, 11),