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 16, 09:30 UTC
[Bugfix] fix detokenizer shallow copy (#5919)
Aurick Qiao committed
Oct 22, 2024 at 18:38 UTC
23b899a8e62c7ea07981bf8487b0dc2cb17847b8
1 file changed
+1
-1
vllm/transformers_utils/detokenizer.py
+1
-1
@@ -90,7 +90,7 @@ class Detokenizer:
90
prefix_offset = next_iter_prefix_offset
91
read_offset = next_iter_read_offset
92
if prev_tokens is None:
93
- prev_tokens = next_iter_tokens
93
+ prev_tokens = next_iter_tokens.copy()
94
else:
95
prev_tokens.extend(next_iter_tokens)
96