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

[Bugfix] fix crash if max_tokens=None (#2570)

Browse Source
This commit is contained in:
Nikola Borisov
2024-01-23 22:38:55 -08:00
committed by GitHub
parent 1e4277d2d1
commit 3209b49033
3 changed files with 28 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/sampling_params.py
View File

@@ -108,7 +108,7 @@ class SamplingParams:
stop_token_ids: Optional[List[int]] = None,
include_stop_str_in_output: bool = False,
ignore_eos: bool = False,
max_tokens: int = 16,
max_tokens: Optional[int] = 16,
logprobs: Optional[int] = None,
prompt_logprobs: Optional[int] = None,
skip_special_tokens: bool = True,
@@ -183,7 +183,7 @@ class SamplingParams:
if not 0.0 <= self.min_p <= 1.0:
raise ValueError("min_p must be in [0, 1], got "
f"{self.min_p}.")
if self.max_tokens < 1:
if self.max_tokens is not None and self.max_tokens < 1:
raise ValueError(
f"max_tokens must be at least 1, got {self.max_tokens}.")
if self.logprobs is not None and self.logprobs < 0: