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 17, 08:00 UTC
fix: bug fix when penalties are negative (#913)
Co-authored-by: dongyong-lee <dongyong.lee@navercorp.com>
Dong-Yong Lee committed
Sep 1, 2023 at 00:37 UTC
e11222333f43c8466c57d0223380dcf297b02bac
1 file changed
+1
-1
vllm/model_executor/layers/sampler.py
+1
-1
@@ -158,7 +158,7 @@ def _apply_penalties(
158
continue
159
p = presence_penalties[i]
160
f = frequency_penalties[i]
161
- if p < _SAMPLING_EPS and f < _SAMPLING_EPS:
161
+ if abs(p) < _SAMPLING_EPS and abs(f) < _SAMPLING_EPS:
162
continue
163
indices.append(i)
164