xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 08:25:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

avoid python list copy in sequence initialization (#401)

Browse Source
This commit is contained in:
Lily Liu 2023-07-08 12:42:08 -07:00 committed by GitHub
parent a945fcc2ae
commit 2179e4f4c5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/sequence.py
View File

@ -127,7 +127,8 @@ class Sequence:
self.logical_token_blocks.append(block) self.logical_token_blocks.append(block)
def _append_tokens_to_blocks(self, token_ids: List[int]) -> None: def _append_tokens_to_blocks(self, token_ids: List[int]) -> None:
while token_ids: cursor = 0
while cursor < len(token_ids):
if not self.logical_token_blocks: if not self.logical_token_blocks:
self._append_logical_block() self._append_logical_block()
@ -137,8 +138,9 @@ class Sequence:
last_block = self.logical_token_blocks[-1] last_block = self.logical_token_blocks[-1]
num_empty_slots = last_block.get_num_empty_slots() num_empty_slots = last_block.get_num_empty_slots()
last_block.append_tokens(token_ids[:num_empty_slots]) last_block.append_tokens(token_ids[cursor:cursor +
token_ids = token_ids[num_empty_slots:] num_empty_slots])
cursor += num_empty_slots
def append_token_id( def append_token_id(
self, self,