biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[v1][KVCacheManager] Rename BlockHashType to BlockHash (#19015)

Browse Source 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
This commit is contained in:
Chen Zhang
2025-06-03 16:01:48 +08:00
committed by GitHub
parent d32aa2e670
commit f32fcd9444
8 changed files with 28 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
tests/v1/core/test_kv_cache_utils.py
View File

@@ -100,8 +100,8 @@ def test_kv_cache_block():
assert block.ref_cnt == 0
# Test block hash setting and resetting
block_hash = vllm.v1.core.kv_cache_utils.BlockHashType(hash_value=123,
token_ids=(1, 2, 3))
block_hash = vllm.v1.core.kv_cache_utils.BlockHash(hash_value=123,
token_ids=(1, 2, 3))
block.block_hash = block_hash
assert block.block_hash == block_hash
@@ -282,7 +282,7 @@ def test_hash_block_tokens(hash_fn):
block_hash = hash_block_tokens(hash_fn, parent_block_hash,
curr_block_token_ids, extra_keys)
assert isinstance(block_hash, vllm.v1.core.kv_cache_utils.BlockHashType)
assert isinstance(block_hash, vllm.v1.core.kv_cache_utils.BlockHash)
assert block_hash.hash_value == hash_fn(
(parent_block_hash, curr_block_token_ids, extra_keys))
assert block_hash.token_ids == curr_block_token_ids
@@ -306,10 +306,8 @@ def test_hash_request_tokens(hash_fn):
block_hashes = hash_request_tokens(hash_fn, block_size, request)
assert len(block_hashes) == 2
assert isinstance(block_hashes[0],
vllm.v1.core.kv_cache_utils.BlockHashType)
assert isinstance(block_hashes[1],
vllm.v1.core.kv_cache_utils.BlockHashType)
assert isinstance(block_hashes[0], vllm.v1.core.kv_cache_utils.BlockHash)
assert isinstance(block_hashes[1], vllm.v1.core.kv_cache_utils.BlockHash)
# Check the first block
assert block_hashes[0].token_ids == (0, 1, 2)

4
tests/v1/core/test_prefix_caching.py
View File

@@ -12,7 +12,7 @@ from vllm.sampling_params import SamplingParams
from vllm.utils import sha256
from vllm.v1.core.block_pool import BlockPool
from vllm.v1.core.kv_cache_manager import KVCacheManager, Request
from vllm.v1.core.kv_cache_utils import (BlockHashType, KVCacheBlock,
from vllm.v1.core.kv_cache_utils import (BlockHash, KVCacheBlock,
hash_block_tokens)
from vllm.v1.kv_cache_interface import (FullAttentionSpec, KVCacheConfig,
KVCacheGroupSpec, SlidingWindowSpec)
@@ -547,7 +547,7 @@ def test_cache_blocks(hash_fn):
# Test that blocks are cached correctly for 2 full blocks from the start.
blocks = [KVCacheBlock(block_id=i) for i in range(2)]
block_hashes: list[BlockHashType] = []
block_hashes: list[BlockHash] = []
block_pool.cache_full_blocks(
request=req,

4
tests/v1/core/test_specialized_manager.py
View File

@@ -3,7 +3,7 @@
import torch
from vllm.v1.core.block_pool import BlockPool
from vllm.v1.core.kv_cache_utils import BlockHashType, KVCacheBlock
from vllm.v1.core.kv_cache_utils import BlockHash, KVCacheBlock
from vllm.v1.core.single_type_kv_cache_manager import SlidingWindowManager
from vllm.v1.kv_cache_interface import SlidingWindowSpec
@@ -32,7 +32,7 @@ def test_sliding_window_possible_cached_prefix():
def run_one_case(block_is_cached, expect_length):
block_hash_list = [
BlockHashType(i, ()) for i in range(len(block_is_cached))
BlockHash(i, ()) for i in range(len(block_is_cached))
]
block_pool.cached_block_hash_to_block.clear()