xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-23 13:44:28 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add max_num_steps to SamplingParams

Browse Source
This commit is contained in:
Woosuk Kwon 2023-02-24 11:44:40 +00:00
parent b39f149a08
commit 46ce1356f7
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
cacheflow/sampling_params.py
View File

@ -10,6 +10,7 @@ class SamplingParams:
top_p: float = 1.0,
use_beam_search: bool = False,
stop_token_ids: Set[int] = [],
max_num_steps: int = 16, # From OpenAI API.
max_context_len: Optional[int] = None,
) -> None:
assert n >= 1
@ -23,6 +24,7 @@ class SamplingParams:
# Zero temperature means greedy decoding.
assert n == 1
assert top_p == 1.0
assert max_num_steps >= 1
assert max_context_len is None or max_context_len >= 0
self.n = n
@ -30,4 +32,5 @@ class SamplingParams:
self.top_p = top_p
self.use_beam_search = use_beam_search
self.stop_token_ids = stop_token_ids
self.max_num_steps = max_num_steps
self.max_context_len = max_context_len