A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Minor] Fix some random typos in comments (#24009)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Aug 31, 2025 at 16:42 UTC
752d2e1c364e4195093e4f3f2fc33e3ae1840707
3 files changed
+3
-3
vllm/utils/__init__.py
+1
-1
@@ -3290,7 +3290,7 @@ def sha256_cbor_64bit(input) -> int:
3290
return full_hash & ((1 << 64) - 1)
3291
3292
3293
-def get_hash_fn_by_name(hash_fn_name: str) -> Callable:
3293
+def get_hash_fn_by_name(hash_fn_name: str) -> Callable[[Any], int]:
3294
"""Get a hash function by name, or raise an error if
3295
the function is not found.
3296
Args:
vllm/v1/core/sched/scheduler.py
+1
-1
@@ -1207,7 +1207,7 @@ class Scheduler(SchedulerInterface):
1207
# Now that the blocks are ready, actually cache them.
1208
(block_ids, ) = self.kv_cache_manager.get_block_ids(request.request_id)
1209
num_computed_tokens = len(block_ids) * self.block_size
1210
- # Handle the case where num request tokens less then one block.
1210
+ # Handle the case where num request tokens less than one block.
1211
num_computed_tokens = min(num_computed_tokens, request.num_tokens)
1212
if num_computed_tokens == request.num_tokens:
1213
num_computed_tokens -= 1
vllm/v1/core/single_type_kv_cache_manager.py
+1
-1
@@ -47,7 +47,7 @@ class SingleTypeKVCacheManager(ABC):
47
# {req_id: The number of cached blocks for this given request}
48
# This is used to track the number of cached blocks for each request.
49
# This is only used to track the RUNNING requests, we do not track the
50
- # data for reempted ones.
50
+ # data for preempted ones.
51
self.num_cached_block: dict[str, int] = {}
52
53
self.kv_cache_group_id = kv_cache_group_id