karylab_agents/vllm
1
0
Fork 0
forked from Karylab-cklius/vllm
Code Pull Requests 2 Actions 2 Packages Activity

[KV Connector] MooncakeStore: drop dead discard_partial_chunks parameter (#43627)

Browse Source 
Signed-off-by: Zhewen Li <zhewen@inferact.ai>
Co-authored-by: Zhewen Li <zhewen@inferact.ai>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Zhewen Li
2026-05-26 13:40:21 -07:00
committed by GitHub
co-authored by Zhewen Li Claude Opus 4.7
parent 6e503868ca
commit d98cbf472b
1 changed files with 2 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/distributed/kv_transfer/kv_connector/v1/mooncake/store/data.py
+2 -11
View File
@@ -223,7 +223,6 @@ class ReqMeta:
skip_save: bool | None = False,
block_hashes: list[BlockHash] | None = None,
is_last_chunk: bool | None = None,
discard_partial_chunks: bool = True,
original_block_size: int | None = None,
) -> "ReqMeta | None":
"""Create ReqMeta from a RequestTracker."""
@@ -231,16 +230,8 @@ class ReqMeta:
block_hashes = []
input_token_len = tracker.token_len
chunk_boundary = (
cdiv(tracker.num_saved_tokens + 1, block_size) * block_size
if discard_partial_chunks
else 0
)
num_tokens_to_save = (
(input_token_len // block_size * block_size)
if discard_partial_chunks
else input_token_len
)
chunk_boundary = cdiv(tracker.num_saved_tokens + 1, block_size) * block_size
num_tokens_to_save = input_token_len // block_size * block_size
skip_save = skip_save or num_tokens_to_save < chunk_boundary
# A ReqMeta must never carry both a save AND a load.