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

[Misc] fix docstrings (#4191)

Browse Source 
Co-authored-by: Zhong Wang <wangzhong@infini-ai.com>
This commit is contained in:
Uranus 2024-04-19 16:18:33 +08:00 committed by GitHub
parent 221d93ecbf
commit 8f20fc04bf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/sequence.py
View File

@ -160,7 +160,7 @@ class SequenceData:
self._stage = SequenceStage.PREFILL self._stage = SequenceStage.PREFILL
def get_num_uncomputed_tokens(self) -> int: def get_num_uncomputed_tokens(self) -> int:
"""Return the number of prefil tokens that are not computed.""" """Return the number of prefill tokens that are not computed."""
# we use `get_len()` which includes prompt_len + output_len instead # we use `get_len()` which includes prompt_len + output_len instead
# of prompt_len here. This is because during recompute we need to # of prompt_len here. This is because during recompute we need to
# prefill for both prompt and output. # prefill for both prompt and output.
@ -345,12 +345,9 @@ class Sequence:
def get_num_new_tokens(self) -> int: def get_num_new_tokens(self) -> int:
"""Get the number of new tokens to be computed. """Get the number of new tokens to be computed.
Args:
remainig_token_budget: The remaining token budgets.
Returns: Returns:
The new number of tokens to be computed. I.e., 1 for decode, prompt The new number of tokens to be computed. I.e., 1 for decode, or
size for prefill. If there's not enough remainig_token_budget, it the remaining prompt size for prefill.
can return the chunked number of new tokens.
""" """
if self.data.stage == SequenceStage.DECODE: if self.data.stage == SequenceStage.DECODE:
return 1 return 1